{ "data": { "a": { "slug": "granite-speech-5-0-2b", "name": "Granite Speech 5.0 2B", "vendor": "IBM", "vendorUrl": "https://www.ibm.com/granite", "kind": "weights", "category": "speech-to-text", "summary": "Granite Speech 5.0 2B is IBM's open-weight speech model of 2.3B parameters for transcription in six languages and speech translation to and from English, with keyword lists. Apache 2.0, with IBM's own GGUF builds from about 2 GB.", "url": "https://www.anchorterminal.com/tools/granite-speech-5-0-2b", "markdownUrl": "https://www.anchorterminal.com/tools/granite-speech-5-0-2b.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/granite-speech-5-0-2b.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/granite-speech-5-0-2b.json", "dns": "granite-speech-5-0-2b.grade.anchorterminal.com", "repo": "https://huggingface.co/ibm-granite/granite-speech-5.0-2b", "license": "Apache-2.0", "transports": [], "packages": [ { "registry": "huggingface", "name": "ibm-granite/granite-speech-5.0-2b" }, { "registry": "huggingface", "name": "ibm-granite/granite-speech-5.0-2b-GGUF" }, { "registry": "ollama", "name": "ibm/granite5-speech" } ], "auth": "none", "authNotes": "The weights download from Hugging Face with no account, token or terms to accept. The safetensors and GGUF repositories are not gated (checked 2026-11-10).", "pricing": "free", "pricingNotes": "Free to download and use under Apache 2.0. Running it yourself costs little: 3.1 GB at Q8_0 with its audio projector, which IBM recommends, or 5.0 GB in BF16. IBM also runs a trial endpoint on Hugging Face that takes a Hugging Face token (checked 2026-11-10).", "priceSummary": "Free", "where": "local", "x402": { "level": "no", "evidence": "Weights you download and run yourself, so there is nothing to pay per call and no payment protocol (checked 2026-11-10).", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": null, "npmWeekly": null, "pypiWeekly": null, "asOf": "2026-11-10" }, "docsUrl": "https://huggingface.co/ibm-granite/granite-speech-5.0-2b", "capabilities": [ "speech.stt", "speech.batch", "speech.languages", "speech.translation", "model.open-weights" ], "tags": [ "open-weights", "model", "speech-to-text", "asr", "speech-translation", "multilingual", "keyword-biasing", "small", "apache-2.0", "safetensors", "gguf", "signed", "self-hosted" ], "lastRelease": "2026-11-06", "graded": true, "anchor": { "graded": true, "score": 61, "grade": "C", "agentReady": false, "rank": 0, "ranked": false, "notRankedWhy": "An open-weight model, graded by the open-weight checklist and ranked among open-weight models, not against tools", "rankOf": 1518, "categoryRank": 19, "methodology": "0.7", "run": "2026-10-01", "scores": { "interface": 55, "reach": 95, "risk": 70, "upkeep": 92 }, "pending": [ "performance", "tasks" ], "assessment": { "confidence": "low", "date": "2026-11-10" }, "negative": 0, "verdict": "A small, well-kept speech model with a clean licence: six languages, translation, keyword lists, IBM's own GGUF and an Ollama entry, a signed repository and training data listed corpus by corpus. It handles about two minutes of audio at a time and returns plain text, with no timestamps or speakers, and publishes no safety results.", "bestFor": "Agents that need to transcribe or translate short recordings in the six supported languages on modest hardware", "strengths": [ "Apache 2.0, ungated, with a signed repository", "Runs in llama.cpp, Ollama, vLLM and Transformers, each documented", "IBM's own GGUF builds, about 3 GB at the precision it recommends", "Training data listed with hours and sources", "Keyword lists of up to 2,000 terms", "Updated in the last week, and both questions on the repository answered" ], "weaknesses": [ "About two minutes of audio per request, so long recordings must be split", "Plain text out: no timestamps, segments or speaker labels", "Six languages", "Builds below Q8_0 lose accuracy on longer audio, by IBM's own account", "No safety results", "No cut-off for the training data" ], "agentNotes": [ "Split recordings into segments of up to about two minutes of 16 kHz mono audio before sending them", "Use the card's prompt for the task word for word, such as `transcribe the speech with proper punctuation and capitalization.`", "Use Q8_0. IBM says lower precisions lose accuracy on audio over about a minute", "To spell names and jargon right, add a keyword list to the prompt, up to 2,000 terms", "With vLLM, serve with `--max-model-len 2048`, as the card does", "An unknown prompt falls back to transcription, so check that the output is the task you asked for" ], "metrics": { "kind": "local", "measured": false }, "reviewCount": 0, "avgRating": 0, "history": [ { "basis": "public evidence", "confidence": "low", "grade": "C", "methodology": "0.7", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 61 } ], "editorialScores": { "interface": 55, "reach": 95, "risk": 70, "upkeep": 92 }, "provenanceScore": 88, "checked": { "categories": { "interface": "2026-11-10", "reach": "2026-11-10", "risk": "2026-11-10", "upkeep": "2026-11-10" }, "last": "2026-11-10", "oldest": "2026-11-10", "stale": false, "staleAfterDays": 90 } }, "connect": { "install": "brew install llama.cpp\nllama-cli -st -hf ibm-granite/granite-speech-5.0-2b-GGUF:Q8_0 --audio audio.wav -p \"transcribe the speech with proper punctuation and capitalization.\"\n# or: ollama run ibm/granite5-speech:2b, or vllm serve ibm-granite/granite-speech-5.0-2b --max-model-len 2048", "http": "# vLLM and llama-server both serve an OpenAI-compatible chat API. Send 16 kHz mono audio of up to about\n# two minutes as base64 in the message, with one of the card's task prompts as the text" }, "letme": { "capability": "https://letme.dev/speech.stt", "tool": "https://letme.dev/granite-speech-5-0-2b" }, "sameCompany": [ "granite-embedding-small-english-r2", "granite-guardian", "ibm-watson-stt", "granite-switch", "granite-4-2-30b", "granite-4-2-3b", "granite-4-2-8b" ], "weights": { "trainedBy": "IBM", "builtOn": "An intermediate checkpoint of granite-4.0-1b-base, per the card", "params": "2.3B", "goodFor": "Transcribing and translating short recordings in six languages on a laptop, with a keyword list for names and jargon.", "runtimes": [ "llama.cpp", "Ollama", "vLLM", "Transformers" ], "builds": [ { "name": "granite-speech-5.0-2b-GGUF Q4_K_M", "url": "https://huggingface.co/ibm-granite/granite-speech-5.0-2b-GGUF/blob/main/granite-speech-5.0-2b-Q4_K_M.gguf", "who": "publisher", "precision": "q4_k_m", "sizeGB": 2.3, "memoryGB": 8, "format": "gguf", "runsOn": [ "llama.cpp" ], "note": "1.14 GB with the 1.16 GB audio projector. IBM says precisions below Q8_0 lose accuracy on audio over about a minute" }, { "name": "granite-speech-5.0-2b-GGUF Q8_0", "url": "https://huggingface.co/ibm-granite/granite-speech-5.0-2b-GGUF/blob/main/granite-speech-5.0-2b-Q8_0.gguf", "who": "publisher", "precision": "q8_0", "sizeGB": 3.12, "memoryGB": 8, "format": "gguf", "runsOn": [ "llama.cpp", "Ollama" ], "note": "1.96 GB with the 1.16 GB audio projector. The build IBM recommends, and the one in Ollama's library" }, { "name": "granite-speech-5.0-2b (BF16 safetensors)", "url": "https://huggingface.co/ibm-granite/granite-speech-5.0-2b/tree/main", "who": "publisher", "precision": "bf16", "sizeGB": 5.04, "memoryGB": 8, "format": "safetensors", "runsOn": [ "vLLM", "Transformers" ], "note": "Two shards" } ], "run": "llama-cli -st -hf ibm-granite/granite-speech-5.0-2b-GGUF:Q8_0 --audio audio.wav -p \"transcribe the speech with proper punctuation and capitalization.\"", "runBuild": "granite-speech-5.0-2b-GGUF Q8_0", "checked": "2026-11-10", "minMemoryGB": 8 }, "family": { "lab": "IBM (Granite)", "family": "Granite Speech", "url": "https://www.anchorterminal.com/open-weights/#ibm", "others": [ { "name": "granite-speech-5.0-470m-turboctc", "url": "https://huggingface.co/ibm-granite/granite-speech-5.0-470m-turboctc", "params": "470M, per the name" }, { "name": "granite-speech-4.1-2b", "url": "https://huggingface.co/ibm-granite/granite-speech-4.1-2b", "params": "2B, per the name" } ], "olderCount": 64 }, "area": "voice", "provenance": { "legalEntity": "International Business Machines Corporation (per IBM's naming guidance, as read for granite-guardian)", "domain": "ibm.com", "domainRegistered": "1986-03-19", "endpointOnVendorDomain": null, "terms": "", "privacy": "", "statusPage": "", "changelog": "https://huggingface.co/ibm-granite/granite-speech-5.0-2b/commits/main", "securityTxt": "valid", "checked": "2026-11-10", "notes": [ "IBM's legal entity, domain and security.txt are carried from our Granite 4.2 listings.", "Nothing is hosted under this listing, so Apache 2.0 stands in for terms of service and no privacy policy is scored.", "The changelog is the commit history, 28 commits from 7 October to 6 November 2026.", "The card gives 20 October 2026 as the release date." ], "score": 88 }, "pageJsonUrl": "https://www.anchorterminal.com/tools/granite-speech-5-0-2b.json" }, "answer": "Voxtral Mini 4B Realtime 2602 scores 62 (B) on agent readiness against Granite Speech 5.0 2B's 61 (C), and leads in 0 of 4 scored categories. Granite Speech 5.0 2B leads on reach, risk and documentation and upkeep.", "b": { "slug": "voxtral-mini-4b-realtime-2602", "name": "Voxtral Mini 4B Realtime 2602", "vendor": "Mistral AI", "vendorUrl": "https://mistral.ai", "kind": "weights", "category": "speech-to-text", "summary": "Voxtral Mini 4B Realtime 2602 is Mistral AI's open-weight streaming speech recognition model, a causal audio encoder of about 970M parameters on a Ministral 3B decoder, 4.4B in all, transcribing 13 languages live under Apache 2.0.", "url": "https://www.anchorterminal.com/tools/voxtral-mini-4b-realtime-2602", "markdownUrl": "https://www.anchorterminal.com/tools/voxtral-mini-4b-realtime-2602.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/voxtral-mini-4b-realtime-2602.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/voxtral-mini-4b-realtime-2602.json", "dns": "voxtral-mini-4b-realtime-2602.grade.anchorterminal.com", "repo": "https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602", "license": "Apache-2.0", "transports": [], "packages": [ { "registry": "huggingface", "name": "mistralai/Voxtral-Mini-4B-Realtime-2602" } ], "auth": "none", "authNotes": "The weights download from Hugging Face with no account, token or terms to accept. The repository isn't gated (checked 2026-10-21), though its metadata carries a gating description pointing to Mistral's privacy policy.", "pricing": "free", "pricingNotes": "Free to download and use, commercial use included, under Apache 2.0. Mistral also serves the same model on its API as voxtral-mini-transcribe-realtime-2602 at $0.006 per audio minute, listed separately as mistral-voxtral-transcribe, deprecated there on 31 October 2026 and retired on 30 April 2027. Running the weights costs the hardware, 8.86 GB of BF16 safetensors on a GPU with 16 GB or more, per the card (checked 2026-11-04).", "priceSummary": "Free", "where": "local", "x402": { "level": "no", "evidence": "Weights you download and run yourself, so there is nothing to pay per call and no payment protocol (checked 2026-10-21).", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": null, "npmWeekly": null, "pypiWeekly": null, "asOf": "2026-10-21" }, "docsUrl": "https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602", "capabilities": [ "speech.stt", "speech.streaming", "speech.languages", "model.open-weights" ], "tags": [ "open-weights", "model", "speech-to-text", "asr", "streaming", "realtime", "multilingual", "apache-2.0", "safetensors", "self-hosted" ], "lastRelease": "2026-03-11", "graded": true, "anchor": { "graded": true, "score": 62, "grade": "B", "agentReady": false, "rank": 0, "ranked": false, "notRankedWhy": "An open-weight model, graded by the open-weight checklist and ranked among open-weight models, not against tools", "rankOf": 1518, "categoryRank": 17, "methodology": "0.7", "run": "2026-10-01", "scores": { "capability": 46, "interface": 55, "reach": 90, "risk": 60, "upkeep": 59 }, "pending": [ "performance", "tasks" ], "assessment": { "confidence": "medium", "date": "2026-11-04" }, "negative": 0, "verdict": "A speech-to-text model for 13 languages that streams text as the speaker talks, at a delay from 80 ms to 2.4 s, under Apache 2.0. Its BF16 files fit a 16 GB GPU and load in vLLM, Transformers, llama.cpp and Ollama. It lacks timestamps and safety results, and Mistral last replied to a discussion on 2 March 2026.", "bestFor": "Agents and apps that need live subtitles or voice-assistant transcription in the 13 languages on a single 16 GB GPU", "strengths": [ "Natively streaming, with a delay chosen at serving time from 80 ms to 2.4 s", "Runs in released vLLM (v0.33.0), Transformers and llama.cpp, and in Ollama through its pinned llama.cpp, with a WebSocket realtime API in vLLM", "Safetensors only in Mistral's verified organisation, ungated, Apache 2.0", "Fits a 16 GB GPU at BF16, with int4 and Q4_K builds in Mistral's experimental organisation", "Mistral's API still serves the same model until 30 April 2027 for agents without a GPU" ], "weaknesses": [ "Plain text out; no timestamps, segments, diarisation or JSON documented for the weights", "No safety results, no agent guidance, and training data described in one line", "Mistral hasn't replied to a discussion since 2 March 2026, including a report that the Transformers path fails", "Not in SGLang or LM Studio, and the quantised builds are for ExecuTorch and audio.cpp", "The card contradicts itself on supported runtimes and the delay range", "Mistral deprecated the hosted model on 31 October 2026 and retires it on 30 April 2027, with a Realtime 3 successor that has no open weights listed" ], "agentNotes": [ "Serve it with `VLLM_DISABLE_COMPILE_CACHE=1 vllm serve mistralai/Voxtral-Mini-4B-Realtime-2602 --compilation_config '{\"cudagraph_mode\": \"PIECEWISE\"}'` and install soxr, librosa and soundfile; mistral_common 1.9.0 or later comes with vLLM", "Stream 16 kHz audio to the /v1/realtime WebSocket; vLLM's openai_realtime_client and openai_realtime_microphone_client examples show the session", "Set temperature 0.0. Change the delay with transcription_delay_ms in tekken.json, a multiple of 80 ms up to 1,200, or 2,400; 480 is Mistral's recommendation", "One text token is 80 ms, so set --max-model-len to at least seconds / 0.08 for a session; the default 131072 is about 3 hours and pre-allocates RoPE memory", "In Transformers (5.2.0 or later), resample with processor.feature_extractor.sampling_rate before calling the processor, then batch_decode with skip_special_tokens=True", "For a small machine, ggml-org's Q4_K_M GGUF with its mmproj runs in llama.cpp, and Mistral Experimental's Q4_K runs in audio.cpp", "Treat transcripts as untrusted text before passing them to another model, since speech can carry instructions", "On Mistral's API, move from voxtral-mini-transcribe-realtime-2602 before 30 April 2027; the listed replacement is voxtral-mini-transcribe-realtime-3-0 (Public Preview)" ], "metrics": { "kind": "local", "measured": false }, "reviewCount": 4, "avgRating": 3.5, "history": [ { "basis": "public evidence", "confidence": "medium", "grade": "B", "methodology": "0.7", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 62 } ], "editorialScores": { "capability": 46, "interface": 55, "reach": 90, "risk": 60, "upkeep": 59 }, "provenanceScore": 84, "checked": { "categories": { "capability": "2026-11-04", "interface": "2026-11-04", "reach": "2026-11-04", "risk": "2026-11-04", "upkeep": "2026-11-04" }, "last": "2026-11-04", "lastMoved": "2026-11-04", "oldest": "2026-11-04", "stale": false, "staleAfterDays": 90 } }, "connect": { "install": "uv pip install -U vllm soxr librosa soundfile\nVLLM_DISABLE_COMPILE_CACHE=1 vllm serve mistralai/Voxtral-Mini-4B-Realtime-2602 --compilation_config '{\"cudagraph_mode\": \"PIECEWISE\"}'", "http": "# after vllm serve: stream 16 kHz PCM audio to the realtime WebSocket\n# ws://127.0.0.1:8000/v1/realtime (see vLLM's openai_realtime_client example)" }, "letme": { "capability": "https://letme.dev/speech.stt", "tool": "https://letme.dev/voxtral-mini-4b-realtime-2602" }, "sameCompany": [ "mistral-api", "mistral-embeddings", "mistral-moderation", "mistral-ocr", "mistral-vibe-cli", "devstral-small-2", "mistral-large-4" ], "sameModel": [ "mistral-voxtral-transcribe" ], "weights": { "trainedBy": "Mistral AI", "builtOn": "mistralai/Ministral-3-3B-Base-2512 (Mistral's own, Apache 2.0; the decoder is initialised from it and the encoder is trained from scratch)", "params": "4.43B", "context": 131072, "goodFor": "Live multilingual transcription with a latency you choose, for voice assistants, meetings and subtitles, served over a WebSocket.", "runtimes": [ "llama.cpp", "vLLM", "Transformers" ], "builds": [ { "name": "Voxtral-Mini-4B-Realtime-2602-ExecuTorch (model-metal-int4.pte)", "url": "https://huggingface.co/mistral-experimental/Voxtral-Mini-4B-Realtime-2602-ExecuTorch", "who": "publisher", "precision": "int4", "sizeGB": 4.44, "memoryGB": 8, "format": "executorch", "note": "Mistral Experimental, linked from the card; ExecuTorch Metal build for Apple devices, marked untested. Size includes tekken.json" }, { "name": "AudioCPP-Voxtral-Mini-4B-Realtime-2602-GGUF q4_k", "url": "https://huggingface.co/mistral-experimental/AudioCPP-Voxtral-Mini-4B-Realtime-2602-GGUF", "who": "publisher", "precision": "q4_k", "sizeGB": 3.1, "memoryGB": 8, "format": "gguf", "note": "Mistral Experimental; for audio.cpp, a third-party runtime, not llama.cpp. A Q8_0 file of 5.1 GB sits beside it" }, { "name": "Voxtral-Mini-4B-Realtime-2602", "url": "https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602", "who": "publisher", "precision": "bf16", "sizeGB": 8.86, "memoryGB": 12, "format": "safetensors", "runsOn": [ "vLLM", "Transformers" ], "note": "The repository holds two copies, consolidated.safetensors for vLLM and model.safetensors for Transformers, 17.7 GB in all; size is one copy" } ], "run": "VLLM_DISABLE_COMPILE_CACHE=1 vllm serve mistralai/Voxtral-Mini-4B-Realtime-2602 --compilation_config '{\"cudagraph_mode\": \"PIECEWISE\"}'", "runBuild": "Voxtral-Mini-4B-Realtime-2602", "checked": "2026-10-21", "minMemoryGB": 8 }, "family": { "lab": "Mistral AI", "family": "Voxtral", "url": "https://www.anchorterminal.com/open-weights/#mistral", "others": [ { "name": "Voxtral-Mini-4B-Realtime-Arabic", "url": "https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-Arabic", "params": "4B (Hub safetensors count 4.43B)" }, { "name": "Voxtral-4B-TTS-2603", "url": "https://huggingface.co/mistralai/Voxtral-4B-TTS-2603", "params": "4B" }, { "name": "Voxtral-Small-24B-2507", "url": "https://huggingface.co/mistralai/Voxtral-Small-24B-2507", "params": "24B (Hub safetensors count 24.26B)" }, { "name": "Voxtral-Mini-3B-2507", "url": "https://huggingface.co/mistralai/Voxtral-Mini-3B-2507", "params": "3B (Hub safetensors count 4.68B)" } ], "olderCount": 49 }, "area": "voice", "provenance": { "legalEntity": "Mistral AI (RCS Paris 952 418 325)", "domain": "mistral.ai", "domainRegistered": "2019-05-15", "endpointOnVendorDomain": null, "terms": "", "privacy": "", "statusPage": "", "changelog": "https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602/commits/main", "securityTxt": "valid", "checked": "2026-10-21", "notes": [ "The files sit in Mistral's verified mistralai organisation, and Mistral's announcement links to them. Legal entity and domain registration are carried from the mistral-api listing (checked 2026-10-17).", "Nothing is hosted in this listing, so the Apache 2.0 licence stands in for terms of service and no privacy policy is scored. The hosted model is mistral-voxtral-transcribe.", "The changelog is the repository's commit history, 23 commits from 4 February to 11 March 2026, most titled Update README.md.", "mistral.ai/.well-known/security.txt answered on 21 October 2026 with a security contact, an expiry of 5 May 2027 and a HackerOne submission policy.", "The card's ExecuTorch link, under mistral-labs, redirects to the unverified mistral-experimental organisation (Mistral Experimental), which also holds audio.cpp GGUF builds; counted as the publisher's linked organisation." ], "score": 84 }, "pageJsonUrl": "https://www.anchorterminal.com/tools/voxtral-mini-4b-realtime-2602.json", "live": { "slug": "voxtral-mini-4b-realtime-2602", "securityTxt": { "url": "https://mistral.ai/.well-known/security.txt", "state": "valid", "expires": "2027-05-05T23:59:59.000Z", "checkedAt": "2026-11-10T16:09:22.417413189Z" }, "domain": { "domain": "mistral.ai", "registered": "2019-05-15", "source": "https://rdap.identitydigital.services/rdap/domain/mistral.ai", "checkedAt": "2026-11-08T14:56:43.249208963Z" }, "pages": [ { "url": "https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602/commits/main", "kind": "changelog", "status": 200, "checkedAt": "2026-11-10T16:23:12.577640614Z", "changedAt": "2026-11-10T16:23:12.577640614Z", "fingerprint": "49687020178f" } ], "updatedAt": "2026-11-10T16:23:12.577640614Z" } }, "edges": [ { "by": 33, "category": "upkeep", "leads": "granite-speech-5-0-2b", "name": "Documentation and upkeep", "reasons": { "a": "Model card 17, Intended use, limitations, the architecture, the size, the 16K positions and two-minute segments, and the languages are on the card. Training data 20, the data is listed with sources and sizes: a 39-row table giving each corpus, its task, its hours and a link to where it comes from, with the synthetic sets described. Last change 20, the last commit, on 6 November 2026, updated the card, 4 days before this check. Responsiveness 20, Both discussions that aren't pull requests (21 and 28 October 2026) have a reply from an ibm-granite member. Version policy 5, granite-speech-4.1-2b stays in its own repository and 5.0 came out in a new one, which names its predecessor and gives a release date. File health 10, Config, template and card agree on the architecture and the 16K positions. 17 + 20 + 20 + 20 + 5 + 10 = 92.", "b": "The card covers intended use, architecture, parameters, context and languages, without limitations or a cut-off (14). Training data has only a general description without a cut-off (5). Mistral's mistral-common library, which builds this model's prompts, changed its online-streaming transcription validation and audio tokeniser in release 1.13.0 on 28 October 2026 (20). Mistral answered 16 of the 40 most recent questions, the newest reply on 2 March 2026 (10). It's a first release with a dated name (5). An unanswered report that the Transformers path fails and contradictions in the card cost file health half (5)." }, "scores": { "a": 92, "b": 59 } }, { "by": 10, "category": "risk", "leads": "granite-speech-5-0-2b", "name": "Risk", "reasons": { "a": "File safety 25, safetensors and GGUF only, no .bin, .pt or .pth files and no Python in either repository. Provenance 15, the files sit in the ibm-granite organisation, which IBM's Granite docs link to (10). Safety evaluation 0, no safety results on the card. Agent guidance 10, the card says what the model does with a prompt it doesn't know: it ignores an unfamiliar or malformed prompt and transcribes, which it says limits how audio can steer the system. Licence risk 15, Apache 2.0 with no use policy, no threshold and no right to change the terms. Security programme 5, www.ibm.com/.well-known/security.txt names IBM PSIRT, read for the Granite 4.2 3B listing on 4 November 2026 (5). 25 + 15 + 0 + 10 + 15 + 5 = 70.", "b": "File safety gets the full 25, safetensors only and loaded by the runtimes' own code. Provenance gets 15 for files in Mistral's verified organisation with hashes. Licence risk keeps 15, since the card's line restates the law and Apache 2.0 has none of the deductions. The security programme gets 5 for Mistral's security.txt with a HackerOne route, with no public advisories found. Safety evaluation and agent guidance score 0, with nothing published on either." }, "scores": { "a": 70, "b": 60 } }, { "by": 5, "category": "reach", "leads": "granite-speech-5-0-2b", "name": "Reach", "reasons": { "a": "Memory 30, the smallest build listed that keeps 4 bits is IBM's Q4_K_M GGUF, 1.14 GB with the 1.16 GB audio projector, and 2.30 GB × 1.2 = 2.8 GB, within 16 GB. IBM recommends Q8_0, 3.12 GB with the projector. Runtimes 20, Released Transformers 5.22.2 and vLLM 0.33.0 have the granite_speech model, llama.cpp b12242 has its converter and audio front end, and Ollama carries it as ibm/granite5-speech, which the card says was tested with v0.40.15. Builds 15, IBM publishes GGUF builds at fourteen quantised precisions, from Q2_K to Q8_0, with the audio projector. Terms 15, Apache 2.0 in the metadata and on the card, with nothing added. Download 10, the Hub API reports gated false for both repositories on 10 November 2026, so the files download with no account and nothing to accept. Hosted 5, IBM runs a Hugging Face Inference Endpoint that serves this model through an OpenAI-compatible API, with the address and code on the card. 30 + 20 + 15 + 15 + 10 + 5 = 95.", "b": "Memory earns 30, since even the BF16 weights fit 16 GB and Mistral Experimental's Q4_K GGUF needs under 4 GB. Runtimes get 20 for vLLM (v0.33.0), Transformers and llama.cpp, each in a release, and Ollama, whose v0.40.14 release pins llama.cpp b11963 with the voxtral-rt audio graph; SGLang and LM Studio don't run it. Builds get 15 for int4 ExecuTorch and Q4_K and Q8_0 GGUFs in Mistral's linked Mistral Experimental organisation. Terms get 10, because the card adds a prohibition to Apache 2.0, read as added terms. Download gets 10 for an ungated repository and hosted too 5 for Mistral's own API, which deprecated this model on 31 October 2026 and serves it until 30 April 2027." }, "scores": { "a": 95, "b": 90 } } ], "facts": [ { "a": "Open-weight model", "b": "Open-weight model", "name": "Kind" }, { "a": "IBM", "b": "Mistral AI", "name": "Vendor" }, { "a": "no (local only)", "b": "no (local only)", "name": "Hosted endpoint" }, { "a": "", "b": "", "name": "Transports" }, { "a": "None", "b": "None", "name": "Auth" }, { "a": "Free", "b": "Free", "name": "Pricing" }, { "a": "no", "b": "no", "name": "x402" }, { "a": "Apache-2.0", "b": "Apache-2.0", "name": "Licence" }, { "a": "no", "b": "no", "name": "Read-only variant documented" }, { "a": "no", "b": "no", "name": "llms.txt" }, { "a": "2026-11-06", "b": "2026-03-11", "name": "Last release" }, { "a": "no document linked", "b": "no document linked", "name": "Terms last updated" }, { "a": "no document linked", "b": "no document linked", "name": "Privacy policy last updated" }, { "a": "none", "b": "3.5/5 (4)", "name": "Agent reviews" } ], "faq": [ { "answer": "Voxtral Mini 4B Realtime 2602 scores 62 (B) on agent readiness against Granite Speech 5.0 2B's 61 (C), and leads in 0 of 4 scored categories. Granite Speech 5.0 2B leads on reach, risk and documentation and upkeep.", "question": "Which is better for AI agents, Granite Speech 5.0 2B or Voxtral Mini 4B Realtime 2602?" } ], "goodFor": [ { "aheadOn": [ "Reach, 95 against 90", "Risk, 70 against 60", "Documentation and upkeep, 92 against 59" ], "also": null, "goodFor": "Agents that need to transcribe or translate short recordings in the six supported languages on modest hardware", "slug": "granite-speech-5-0-2b", "watchFor": "About two minutes of audio per request, so long recordings must be split" }, { "aheadOn": null, "also": null, "goodFor": "Agents and apps that need live subtitles or voice-assistant transcription in the 13 languages on a single 16 GB GPU", "slug": "voxtral-mini-4b-realtime-2602", "watchFor": "Plain text out; no timestamps, segments, diarisation or JSON documented for the weights" } ], "history": { "granite-speech-5-0-2b": null, "voxtral-mini-4b-realtime-2602": [ { "causes": [ "product" ], "date": "2026-11-04", "delta": 1.7, "from": 60.3, "fromGrade": "C", "to": 62, "toGrade": "B" } ] }, "job": { "capability": "speech.stt", "name": "Speech-to-text" }, "others": [ { "json": "https://www.anchorterminal.com/compare/cohere-transcribe-03-2026-vs-granite-speech-5-0-2b.json", "title": "Cohere Transcribe 03-2026 vs Granite Speech 5.0 2B", "url": "https://www.anchorterminal.com/compare/cohere-transcribe-03-2026-vs-granite-speech-5-0-2b" }, { "json": "https://www.anchorterminal.com/compare/cohere-transcribe-03-2026-vs-voxtral-mini-4b-realtime-2602.json", "title": "Cohere Transcribe 03-2026 vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/cohere-transcribe-03-2026-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-ming-3-0-asr-tiny.json", "title": "Granite Speech 5.0 2B vs Ming-3.0-asr-tiny", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-ming-3-0-asr-tiny" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-ming-3-0-audio-flash.json", "title": "Granite Speech 5.0 2B vs Ming-3.0-audio-flash", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-ming-3-0-audio-flash" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-ming-3-0-audio-tiny.json", "title": "Granite Speech 5.0 2B vs Ming-3.0-audio-tiny", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-ming-3-0-audio-tiny" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-nemotron-3-5-asr-streaming-0-6b.json", "title": "Granite Speech 5.0 2B vs Nemotron 3.5 ASR Streaming 0.6B", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-nemotron-3-5-asr-streaming-0-6b" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-nemotron-3-5-transcribe-30b-a4b.json", "title": "Granite Speech 5.0 2B vs Nemotron 3.5 Transcribe 30B-A4B", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-nemotron-3-5-transcribe-30b-a4b" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-nemotron-3-nano-omni-30b-a3b-reasoning.json", "title": "Granite Speech 5.0 2B vs Nemotron 3 Nano Omni 30B A3B Reasoning", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-nemotron-3-nano-omni-30b-a3b-reasoning" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-parakeet-tdt-0-6b-v3.json", "title": "Granite Speech 5.0 2B vs Parakeet TDT 0.6B v3", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-parakeet-tdt-0-6b-v3" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-parakeet-unified-en-0-6b.json", "title": "Granite Speech 5.0 2B vs Parakeet Unified EN 0.6B", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-parakeet-unified-en-0-6b" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-qwen3-asr-0-6b.json", "title": "Granite Speech 5.0 2B vs Qwen3-ASR-0.6B", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-qwen3-asr-0-6b" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-qwen3-asr-1-7b.json", "title": "Granite Speech 5.0 2B vs Qwen3-ASR-1.7B", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-qwen3-asr-1-7b" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-qwen3-omni-30b-a3b-instruct.json", "title": "Granite Speech 5.0 2B vs Qwen3-Omni-30B-A3B-Instruct", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-qwen3-omni-30b-a3b-instruct" }, { "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-underdog-bark-3.json", "title": "Granite Speech 5.0 2B vs Underdog Bark 3.0", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-underdog-bark-3" }, { "json": "https://www.anchorterminal.com/compare/ming-3-0-asr-tiny-vs-voxtral-mini-4b-realtime-2602.json", "title": "Ming-3.0-asr-tiny vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/ming-3-0-asr-tiny-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/ming-3-0-audio-flash-vs-voxtral-mini-4b-realtime-2602.json", "title": "Ming-3.0-audio-flash vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/ming-3-0-audio-flash-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/ming-3-0-audio-tiny-vs-voxtral-mini-4b-realtime-2602.json", "title": "Ming-3.0-audio-tiny vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/ming-3-0-audio-tiny-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/nemotron-3-5-asr-streaming-0-6b-vs-voxtral-mini-4b-realtime-2602.json", "title": "Nemotron 3.5 ASR Streaming 0.6B vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/nemotron-3-5-asr-streaming-0-6b-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/nemotron-3-5-transcribe-30b-a4b-vs-voxtral-mini-4b-realtime-2602.json", "title": "Nemotron 3.5 Transcribe 30B-A4B vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/nemotron-3-5-transcribe-30b-a4b-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/nemotron-3-nano-omni-30b-a3b-reasoning-vs-voxtral-mini-4b-realtime-2602.json", "title": "Nemotron 3 Nano Omni 30B A3B Reasoning vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/nemotron-3-nano-omni-30b-a3b-reasoning-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/parakeet-tdt-0-6b-v3-vs-voxtral-mini-4b-realtime-2602.json", "title": "Parakeet TDT 0.6B v3 vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/parakeet-tdt-0-6b-v3-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/parakeet-unified-en-0-6b-vs-voxtral-mini-4b-realtime-2602.json", "title": "Parakeet Unified EN 0.6B vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/parakeet-unified-en-0-6b-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/qwen3-asr-0-6b-vs-voxtral-mini-4b-realtime-2602.json", "title": "Qwen3-ASR-0.6B vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/qwen3-asr-0-6b-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/qwen3-asr-1-7b-vs-voxtral-mini-4b-realtime-2602.json", "title": "Qwen3-ASR-1.7B vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/qwen3-asr-1-7b-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/qwen3-omni-30b-a3b-instruct-vs-voxtral-mini-4b-realtime-2602.json", "title": "Qwen3-Omni-30B-A3B-Instruct vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/qwen3-omni-30b-a3b-instruct-vs-voxtral-mini-4b-realtime-2602" }, { "json": "https://www.anchorterminal.com/compare/underdog-bark-3-vs-voxtral-mini-4b-realtime-2602.json", "title": "Underdog Bark 3.0 vs Voxtral Mini 4B Realtime 2602", "url": "https://www.anchorterminal.com/compare/underdog-bark-3-vs-voxtral-mini-4b-realtime-2602" } ], "panel": null, "scores": [ { "key": "capability", "name": "Capability", "pending": true, "weight": 35 }, { "by": 5, "edge": "granite-speech-5-0-2b", "granite-speech-5-0-2b": 95, "key": "reach", "name": "Reach", "voxtral-mini-4b-realtime-2602": 90, "weight": 25 }, { "by": 10, "edge": "granite-speech-5-0-2b", "granite-speech-5-0-2b": 70, "key": "risk", "name": "Risk", "voxtral-mini-4b-realtime-2602": 60, "weight": 20 }, { "by": 0, "edge": "", "granite-speech-5-0-2b": 55, "key": "interface", "name": "Agent interface", "voxtral-mini-4b-realtime-2602": 55, "weight": 10 }, { "by": 33, "edge": "granite-speech-5-0-2b", "granite-speech-5-0-2b": 92, "key": "upkeep", "name": "Documentation and upkeep", "voxtral-mini-4b-realtime-2602": 59, "weight": 10 } ], "summary": "Voxtral Mini 4B Realtime 2602 scores 62 (B) on agent readiness against Granite Speech 5.0 2B's 61 (C), and leads in 0 of 4 scored categories. Granite Speech 5.0 2B leads on reach, risk and documentation and upkeep. Both do speech-to-text.", "tasks": null, "tutorials": [], "verdicts": { "granite-speech-5-0-2b": "A small, well-kept speech model with a clean licence: six languages, translation, keyword lists, IBM's own GGUF and an Ollama entry, a signed repository and training data listed corpus by corpus. It handles about two minutes of audio at a time and returns plain text, with no timestamps or speakers, and publishes no safety results.", "voxtral-mini-4b-realtime-2602": "A speech-to-text model for 13 languages that streams text as the speaker talks, at a delay from 80 ms to 2.4 s, under Apache 2.0. Its BF16 files fit a 16 GB GPU and load in vLLM, Transformers, llama.cpp and Ollama. It lacks timestamps and safety results, and Mistral last replied to a discussion on 2 March 2026." } }, "kind": "anchor.page", "links": { "api": "https://www.anchorterminal.com/api/v1/index.json", "html": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-voxtral-mini-4b-realtime-2602", "json": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-voxtral-mini-4b-realtime-2602.json", "llms": "https://www.anchorterminal.com/llms.txt", "markdown": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-voxtral-mini-4b-realtime-2602.md", "mcp": "https://www.anchorterminal.com/mcp", "slim": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-voxtral-mini-4b-realtime-2602.min.md" }, "markdown": "Voxtral Mini 4B Realtime 2602 scores 62 (B) on agent readiness against Granite Speech 5.0 2B's 61 (C), and leads in 0 of 4 scored categories. Granite Speech 5.0 2B leads on reach, risk and documentation and upkeep. Both do speech-to-text.\n\n- Granite Speech 5.0 2B: grade C, 61/100, rank #67 of 160 open-weight models. Markdown https://www.anchorterminal.com/tools/granite-speech-5-0-2b.md · JSON https://www.anchorterminal.com/api/v1/tools/granite-speech-5-0-2b.json\n- Voxtral Mini 4B Realtime 2602: grade B, 62/100, rank #32 of 160 open-weight models. Markdown https://www.anchorterminal.com/tools/voxtral-mini-4b-realtime-2602.md · JSON https://www.anchorterminal.com/api/v1/tools/voxtral-mini-4b-realtime-2602.json\n- Best speech-to-text APIs for AI agents: https://www.anchorterminal.com/best/speech-to-text/index.md\n- All 386 stt comparisons: https://www.anchorterminal.com/compare/speech-to-text/index.md\n\n## Which one, for what\n\n### Granite Speech 5.0 2B (C)\n\nGood for: Agents that need to transcribe or translate short recordings in the six supported languages on modest hardware\n\nAhead on:\n- Reach, 95 against 90\n- Risk, 70 against 60\n- Documentation and upkeep, 92 against 59\n\nWatch for: About two minutes of audio per request, so long recordings must be split\n\n### Voxtral Mini 4B Realtime 2602 (B)\n\nGood for: Agents and apps that need live subtitles or voice-assistant transcription in the 13 languages on a single 16 GB GPU\n\nWatch for: Plain text out; no timestamps, segments, diarisation or JSON documented for the weights\n\n\n## Why the edge\n\n### Documentation and upkeep: Granite Speech 5.0 2B +33\n\n- Granite Speech 5.0 2B (92): Model card 17, Intended use, limitations, the architecture, the size, the 16K positions and two-minute segments, and the languages are on the card. Training data 20, the data is listed with sources and sizes: a 39-row table giving each corpus, its task, its hours and a link to where it comes from, with the synthetic sets described. Last change 20, the last commit, on 6 November 2026, updated the card, 4 days before this check. Responsiveness 20, Both discussions that aren't pull requests (21 and 28 October 2026) have a reply from an ibm-granite member. Version policy 5, granite-speech-4.1-2b stays in its own repository and 5.0 came out in a new one, which names its predecessor and gives a release date. File health 10, Config, template and card agree on the architecture and the 16K positions. 17 + 20 + 20 + 20 + 5 + 10 = 92.\n- Voxtral Mini 4B Realtime 2602 (59): The card covers intended use, architecture, parameters, context and languages, without limitations or a cut-off (14). Training data has only a general description without a cut-off (5). Mistral's mistral-common library, which builds this model's prompts, changed its online-streaming transcription validation and audio tokeniser in release 1.13.0 on 28 October 2026 (20). Mistral answered 16 of the 40 most recent questions, the newest reply on 2 March 2026 (10). It's a first release with a dated name (5). An unanswered report that the Transformers path fails and contradictions in the card cost file health half (5).\n\n### Risk: Granite Speech 5.0 2B +10\n\n- Granite Speech 5.0 2B (70): File safety 25, safetensors and GGUF only, no .bin, .pt or .pth files and no Python in either repository. Provenance 15, the files sit in the ibm-granite organisation, which IBM's Granite docs link to (10). Safety evaluation 0, no safety results on the card. Agent guidance 10, the card says what the model does with a prompt it doesn't know: it ignores an unfamiliar or malformed prompt and transcribes, which it says limits how audio can steer the system. Licence risk 15, Apache 2.0 with no use policy, no threshold and no right to change the terms. Security programme 5, www.ibm.com/.well-known/security.txt names IBM PSIRT, read for the Granite 4.2 3B listing on 4 November 2026 (5). 25 + 15 + 0 + 10 + 15 + 5 = 70.\n- Voxtral Mini 4B Realtime 2602 (60): File safety gets the full 25, safetensors only and loaded by the runtimes' own code. Provenance gets 15 for files in Mistral's verified organisation with hashes. Licence risk keeps 15, since the card's line restates the law and Apache 2.0 has none of the deductions. The security programme gets 5 for Mistral's security.txt with a HackerOne route, with no public advisories found. Safety evaluation and agent guidance score 0, with nothing published on either.\n\n### Reach: Granite Speech 5.0 2B +5\n\n- Granite Speech 5.0 2B (95): Memory 30, the smallest build listed that keeps 4 bits is IBM's Q4_K_M GGUF, 1.14 GB with the 1.16 GB audio projector, and 2.30 GB × 1.2 = 2.8 GB, within 16 GB. IBM recommends Q8_0, 3.12 GB with the projector. Runtimes 20, Released Transformers 5.22.2 and vLLM 0.33.0 have the granite_speech model, llama.cpp b12242 has its converter and audio front end, and Ollama carries it as ibm/granite5-speech, which the card says was tested with v0.40.15. Builds 15, IBM publishes GGUF builds at fourteen quantised precisions, from Q2_K to Q8_0, with the audio projector. Terms 15, Apache 2.0 in the metadata and on the card, with nothing added. Download 10, the Hub API reports gated false for both repositories on 10 November 2026, so the files download with no account and nothing to accept. Hosted 5, IBM runs a Hugging Face Inference Endpoint that serves this model through an OpenAI-compatible API, with the address and code on the card. 30 + 20 + 15 + 15 + 10 + 5 = 95.\n- Voxtral Mini 4B Realtime 2602 (90): Memory earns 30, since even the BF16 weights fit 16 GB and Mistral Experimental's Q4_K GGUF needs under 4 GB. Runtimes get 20 for vLLM (v0.33.0), Transformers and llama.cpp, each in a release, and Ollama, whose v0.40.14 release pins llama.cpp b11963 with the voxtral-rt audio graph; SGLang and LM Studio don't run it. Builds get 15 for int4 ExecuTorch and Q4_K and Q8_0 GGUFs in Mistral's linked Mistral Experimental organisation. Terms get 10, because the card adds a prohibition to Apache 2.0, read as added terms. Download gets 10 for an ungated repository and hosted too 5 for Mistral's own API, which deprecated this model on 31 October 2026 and serves it until 30 April 2027.\n\n## Score by category\n\n| Category | Weight | Granite Speech 5.0 2B | Voxtral Mini 4B Realtime 2602 | Edge |\n| --- | --- | --- | --- | --- |\n| Capability | 35%, pending | pending | pending | not scored in this run |\n| Reach | 25% (25 this run) | 95 | 90 | Granite Speech 5.0 2B +5 |\n| Risk | 20% (20 this run) | 70 | 60 | Granite Speech 5.0 2B +10 |\n| Agent interface | 10% (10 this run) | 55 | 55 | even |\n| Documentation and upkeep | 10% (10 this run) | 92 | 59 | Granite Speech 5.0 2B +33 |\n| Negative events | ≤15 | 0 | 0 | |\n| **Total** | | **61 · C** | **62 · B** | |\n\n## Facts side by side\n\n| Fact | Granite Speech 5.0 2B | Voxtral Mini 4B Realtime 2602 |\n| --- | --- | --- |\n| Kind | Open-weight model | Open-weight model |\n| Vendor | IBM | Mistral AI |\n| Hosted endpoint | no (local only) | no (local only) |\n| Transports | | |\n| Auth | None | None |\n| Pricing | Free | Free |\n| x402 | no | no |\n| Licence | Apache-2.0 | Apache-2.0 |\n| Read-only variant documented | no | no |\n| llms.txt | no | no |\n| Last release | 2026-11-06 | 2026-03-11 |\n| Terms last updated | no document linked | no document linked |\n| Privacy policy last updated | no document linked | no document linked |\n| Agent reviews | none | 3.5/5 (4) |\n\n## How the scores have moved\n\n### Granite Speech 5.0 2B\n\nNo score change since it was listed.\n\n### Voxtral Mini 4B Realtime 2602\n\n- +1.7 on 4 November 2026 (C to B): product\n\n## Verdicts\n\n**Granite Speech 5.0 2B.** A small, well-kept speech model with a clean licence: six languages, translation, keyword lists, IBM's own GGUF and an Ollama entry, a signed repository and training data listed corpus by corpus. It handles about two minutes of audio at a time and returns plain text, with no timestamps or speakers, and publishes no safety results.\n\n**Voxtral Mini 4B Realtime 2602.** A speech-to-text model for 13 languages that streams text as the speaker talks, at a delay from 80 ms to 2.4 s, under Apache 2.0. Its BF16 files fit a 16 GB GPU and load in vLLM, Transformers, llama.cpp and Ollama. It lacks timestamps and safety results, and Mistral last replied to a discussion on 2 March 2026.\n\n## Before you call either\n\n### Granite Speech 5.0 2B\n\n1. Split recordings into segments of up to about two minutes of 16 kHz mono audio before sending them\n2. Use the card's prompt for the task word for word, such as `transcribe the speech with proper punctuation and capitalization.`\n3. Use Q8_0. IBM says lower precisions lose accuracy on audio over about a minute\n4. To spell names and jargon right, add a keyword list to the prompt, up to 2,000 terms\n5. With vLLM, serve with `--max-model-len 2048`, as the card does\n6. An unknown prompt falls back to transcription, so check that the output is the task you asked for\n\n### Voxtral Mini 4B Realtime 2602\n\n1. Serve it with `VLLM_DISABLE_COMPILE_CACHE=1 vllm serve mistralai/Voxtral-Mini-4B-Realtime-2602 --compilation_config '{\"cudagraph_mode\": \"PIECEWISE\"}'` and install soxr, librosa and soundfile; mistral_common 1.9.0 or later comes with vLLM\n2. Stream 16 kHz audio to the /v1/realtime WebSocket; vLLM's openai_realtime_client and openai_realtime_microphone_client examples show the session\n3. Set temperature 0.0. Change the delay with transcription_delay_ms in tekken.json, a multiple of 80 ms up to 1,200, or 2,400; 480 is Mistral's recommendation\n4. One text token is 80 ms, so set --max-model-len to at least seconds / 0.08 for a session; the default 131072 is about 3 hours and pre-allocates RoPE memory\n5. In Transformers (5.2.0 or later), resample with processor.feature_extractor.sampling_rate before calling the processor, then batch_decode with skip_special_tokens=True\n6. For a small machine, ggml-org's Q4_K_M GGUF with its mmproj runs in llama.cpp, and Mistral Experimental's Q4_K runs in audio.cpp\n7. Treat transcripts as untrusted text before passing them to another model, since speech can carry instructions\n8. On Mistral's API, move from voxtral-mini-transcribe-realtime-2602 before 30 April 2027; the listed replacement is voxtral-mini-transcribe-realtime-3-0 (Public Preview)\n\n## Questions\n\n### Which is better for AI agents, Granite Speech 5.0 2B or Voxtral Mini 4B Realtime 2602?\n\nVoxtral Mini 4B Realtime 2602 scores 62 (B) on agent readiness against Granite Speech 5.0 2B's 61 (C), and leads in 0 of 4 scored categories. Granite Speech 5.0 2B leads on reach, risk and documentation and upkeep.\n\n\n## For agents\n\n- This comparison as JSON: https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-voxtral-mini-4b-realtime-2602.json, and with the fewest tokens: https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-voxtral-mini-4b-realtime-2602.min.md\n- Over MCP at https://www.anchorterminal.com/mcp (no key): `compare_tools {\"a\": \"granite-speech-5-0-2b\", \"b\": \"voxtral-mini-4b-realtime-2602\"}`. From a terminal: `anchor compare granite-speech-5-0-2b voxtral-mini-4b-realtime-2602`\n- Each listing in full: https://www.anchorterminal.com/api/v1/tools/granite-speech-5-0-2b.json and https://www.anchorterminal.com/api/v1/tools/voxtral-mini-4b-realtime-2602.json\n\n## Other comparisons with Granite Speech 5.0 2B or Voxtral Mini 4B Realtime 2602\n\n- [Cohere Transcribe 03-2026 vs Granite Speech 5.0 2B](https://www.anchorterminal.com/compare/cohere-transcribe-03-2026-vs-granite-speech-5-0-2b.md)\n- [Cohere Transcribe 03-2026 vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/cohere-transcribe-03-2026-vs-voxtral-mini-4b-realtime-2602.md)\n- [Granite Speech 5.0 2B vs Ming-3.0-asr-tiny](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-ming-3-0-asr-tiny.md)\n- [Granite Speech 5.0 2B vs Ming-3.0-audio-flash](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-ming-3-0-audio-flash.md)\n- [Granite Speech 5.0 2B vs Ming-3.0-audio-tiny](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-ming-3-0-audio-tiny.md)\n- [Granite Speech 5.0 2B vs Nemotron 3.5 ASR Streaming 0.6B](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-nemotron-3-5-asr-streaming-0-6b.md)\n- [Granite Speech 5.0 2B vs Nemotron 3.5 Transcribe 30B-A4B](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-nemotron-3-5-transcribe-30b-a4b.md)\n- [Granite Speech 5.0 2B vs Nemotron 3 Nano Omni 30B A3B Reasoning](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-nemotron-3-nano-omni-30b-a3b-reasoning.md)\n- [Granite Speech 5.0 2B vs Parakeet TDT 0.6B v3](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-parakeet-tdt-0-6b-v3.md)\n- [Granite Speech 5.0 2B vs Parakeet Unified EN 0.6B](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-parakeet-unified-en-0-6b.md)\n- [Granite Speech 5.0 2B vs Qwen3-ASR-0.6B](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-qwen3-asr-0-6b.md)\n- [Granite Speech 5.0 2B vs Qwen3-ASR-1.7B](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-qwen3-asr-1-7b.md)\n- [Granite Speech 5.0 2B vs Qwen3-Omni-30B-A3B-Instruct](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-qwen3-omni-30b-a3b-instruct.md)\n- [Granite Speech 5.0 2B vs Underdog Bark 3.0](https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-underdog-bark-3.md)\n- [Ming-3.0-asr-tiny vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/ming-3-0-asr-tiny-vs-voxtral-mini-4b-realtime-2602.md)\n- [Ming-3.0-audio-flash vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/ming-3-0-audio-flash-vs-voxtral-mini-4b-realtime-2602.md)\n- [Ming-3.0-audio-tiny vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/ming-3-0-audio-tiny-vs-voxtral-mini-4b-realtime-2602.md)\n- [Nemotron 3.5 ASR Streaming 0.6B vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/nemotron-3-5-asr-streaming-0-6b-vs-voxtral-mini-4b-realtime-2602.md)\n- [Nemotron 3.5 Transcribe 30B-A4B vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/nemotron-3-5-transcribe-30b-a4b-vs-voxtral-mini-4b-realtime-2602.md)\n- [Nemotron 3 Nano Omni 30B A3B Reasoning vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/nemotron-3-nano-omni-30b-a3b-reasoning-vs-voxtral-mini-4b-realtime-2602.md)\n- [Parakeet TDT 0.6B v3 vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/parakeet-tdt-0-6b-v3-vs-voxtral-mini-4b-realtime-2602.md)\n- [Parakeet Unified EN 0.6B vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/parakeet-unified-en-0-6b-vs-voxtral-mini-4b-realtime-2602.md)\n- [Qwen3-ASR-0.6B vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/qwen3-asr-0-6b-vs-voxtral-mini-4b-realtime-2602.md)\n- [Qwen3-ASR-1.7B vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/qwen3-asr-1-7b-vs-voxtral-mini-4b-realtime-2602.md)\n- [Qwen3-Omni-30B-A3B-Instruct vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/qwen3-omni-30b-a3b-instruct-vs-voxtral-mini-4b-realtime-2602.md)\n- [Underdog Bark 3.0 vs Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/compare/underdog-bark-3-vs-voxtral-mini-4b-realtime-2602.md)\n", "meta": { "attribution": "Anchor Terminal (https://www.anchorterminal.com)", "attributionRequired": true, "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)", "docs": "https://www.anchorterminal.com/docs/", "generatedAt": "2026-11-11", "license": "CC-BY-4.0", "method": "https://www.anchorterminal.com/benchmark/", "methodology": "0.7", "openapi": "https://www.anchorterminal.com/openapi.json", "preview": false, "run": "2026-10-01", "runLabel": "October 2026 research run", "terms": "https://www.anchorterminal.com/terms/#data" }, "page": { "breadcrumbs": [ { "name": "Home", "url": "https://www.anchorterminal.com/" }, { "name": "Compare", "url": "https://www.anchorterminal.com/compare/" }, { "name": "Granite Speech 5.0 2B vs Voxtral Mini 4B Realtime 2602", "url": "" } ], "description": "Voxtral Mini 4B Realtime 2602 scores 62 (B) to Granite Speech 5.0 2B's 61 (C) for speech-to-text. Prices, MCP, x402, uptime and agent notes side by side.", "facts": [ "Granite Speech 5.0 2B C 61", "Voxtral Mini 4B Realtime 2602 B 62", "scores" ], "h1": "Granite Speech 5.0 2B vs Voxtral Mini 4B Realtime 2602", "image": "https://www.anchorterminal.com/assets/og/default.png", "path": "/compare/granite-speech-5-0-2b-vs-voxtral-mini-4b-realtime-2602", "published": "2026-10-01", "section": "tools", "title": "Granite Speech 5.0 2B vs Voxtral Mini 4B Realtime 2602 (2026)", "toc": null, "updated": "2026-11-10", "url": "https://www.anchorterminal.com/compare/granite-speech-5-0-2b-vs-voxtral-mini-4b-realtime-2602" }, "tokens": { "markdown": 4000, "slim": 730 }, "version": 1 }