Hy-Memory-30B-A3B vs Muse Glimmer 30B

Hy-Memory-30B-A3B 60

File safety 25, 51 safetensors shards and no pickle or Python files. Provenance 15, files in the tencent organisation that the card's own resource list links to (10), with a SHA-256 listed for all 53 large files (5). Safety evaluation 0, no safety results on the card. Agent guidance 0, no prompt-injection guidance, instruction hierarchy, account of trained refusals or general cautions on the card. Licence risk 15, Apache 2.0 with no use policy, no field-of-use limit, no threshold and no clause letting Tencent change the terms. Security programme 5, Tencent's Security Response Center takes vulnerability reports company-wide with published reward rules, as the Hy3 listing records (5). 25 + 15 + 0 + 0 + 15 + 5 = 60.

Muse Glimmer 30B 80

File safety 25, safetensors and GGUF, with the ExecuTorch pos_embed.bin files read as raw embedding tables by their exact size and the Hub's scans, and no repository code runs. Provenance 15, the verified meta-models organisation Meta links, with a checksum on every file. Safety evaluation 10, Meta's own published results with stated set-ups but nothing to rerun and no outside assessor. Agent guidance 15, trained against indirect prompt injection with a principal hierarchy, and results published. Licence risk 10, 5 off for the use policy, and the policy's pointer to a latest copy leads back to the same repository file with no change clause. Security programme 5, a whitehat route in the usage policy, with no security.txt and public advisories unchecked, so scored as absent.

Hy-Memory-30B-A3B 65

Chat template 20, chat_template.jinja in the repository. Tool-calling format 10, the card doesn't mention tools. Trained tool calling 10, the template can render tool calls, but the card presents the model as a memory writer that returns memories as text or JSON and says nothing of calling tools. Structured output 5, the card's quickstart asks for a JSON object with a memories list in the system prompt and shows nothing else, no response_format and no schema, so JSON output is shown by example without a documented way to constrain it. Reasoning control 10, the card passes reasoning_effort as no_think per request through chat_template_kwargs, and the template takes high, low or no_think. Context 10, 262,144 tokens native (max_position_embeddings in config.json). 20 + 10 + 10 + 5 + 10 + 10 = 65.

Muse Glimmer 30B 75

Chat template 20, a Jinja file in the repository. Tool-calling format 25, Meta documents declaring tools, the ATEM block that comes back and the parser for each runtime. Trained tool calling 10, one call per turn and no parallel calls. Structured output 0, nothing documented for JSON or JSON Schema. Reasoning control 10, reasoning can't be turned off but is budgeted per request with four strength levels and returned as reasoning_content. Context 10, 131,072 tokens trained in.

Hy-Memory-30B-A3B 85

Memory 25, Tencent's own Q4_K_M GGUF is 18.24 GB, and 18.24 GB × 1.2 = 21.9 GB, which fits 24 GB but not 16 GB. Its Q3_K_M file and the two files with experts at 1 or 2 bits are below 4-bit precision and don't count. Runtimes 20, hy_v3 model code in released vLLM (HYV3ForCausalLM in the v0.33.0 registry), SGLang v0.5.23 (hunyuan_v3.py), Transformers v5.22.2 and llama.cpp b12242, 5 each. Builds 15, Tencent publishes an FP8 repository and a GGUF repository with Q4_K_M, Q6_K and Q8_0 files in its own organisation, so quantised builds at two precisions or more. Terms 15, Apache 2.0, with only a preamble naming Tencent as copyright holder above the licence text. Download 10, gated is false for the main, FP8 and GGUF repositories on 10 November 2026, so the files download with no account or terms to accept. Hosted 0, Hugging Face lists no inference provider for the model (an empty provider mapping on 10 November 2026), and the card names no API. 25 + 20 + 15 + 15 + 10 + 0 = 85.

Muse Glimmer 30B 90

Memory 25, Meta's 16.8 GB Q4_K_M GGUF needs about 20 GB with the runtime's fifth, so it fits a 24 GB card but not 16 GB. Runtimes 25, released model code in vLLM, SGLang, Transformers, llama.cpp and Ollama's library fills the cap. Builds 15, Meta publishes GGUF and ExecuTorch 4-bit builds, read as two kinds as gemma-4-31b-it was, though all are 4-bit k-quants. Terms 10, Apache 2.0 with a usage policy that adds limits beyond the law, so a free commercial licence under a use policy. Download 10, no repository is gated. Hosted 5, Meta's API doesn't serve it but Together, Featherless AI and DeepInfra are live on its Hub page.