Tencent Hy4 vs Muse Spark Open

Tencent Hy4 45

Model reading for hy4 on Tencent Cloud TokenHub. Platform facts are the tencent-tokenhub listing's (checked 1 November), carried over as RULES.md allows for a model on an already graded platform. No machine payment protocol (0). The price page, updated 9 November, lists ¥6 input, ¥0.3 cached and ¥18 output per million tokens, with Hy4 billed at 50% until 31 December 2026 (20). TokenHub's campaign of 1M free tokens per language model, read on 1 November, isn't confirmed for hy4 on the price page, so half (10 of 20, a judgement call). A person signs up with real-name verification (0). 0 + 20 + 10 + 0 = 30. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 30 is 45.

Muse Spark Open 75

Self-hosted rule. Nothing to buy from Meta, and the licence grants use free of charge. The Model as a Service clause asks for a separate agreement only from licensee groups above $20M gross revenue in 12 months (20). Free to run with no card (20). Meta's own weights need a manually reviewed Hugging Face form with legal name and date of birth, or, since the 4 November re-read of Get the model, a signed-in download from Meta's own page, so an agent can't get them alone. The ungated GGUF quantisations from ggml-org and Unsloth that Meta's docs point to download without approval (the Unsloth repository is ungated on 4 November), so part of the line (10 of 20). No payment protocol (0). Hosted copies at $1.25 input and $4.25 output per million tokens need each provider's own account and are not graded. 20 + 20 + 10 + 0 = 50. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 50 is 75.

Tencent Hy4 71

Model reading for hy4 on Tencent Cloud TokenHub. Platform facts are the tencent-tokenhub listing's (checked 1 November), carried over as RULES.md allows for a model on an already graded platform. Bearer keys limited by model, service, IP range and token quota, long-lived with no rotation (26 of 30). The terms say inputs aren't used to train Tencent Cloud's own models, and Hy4 is first party, so no third party's terms apply (14 of 20). Retention only as long as the service needs, no periods (7 of 15). Per-call audit logs and optional full logs to the customer's CLS (14 of 15). No security.txt; Tencent's TSRC bounty covers Tencent Cloud; no certification checked (10 of 20). 26 + 14 + 7 + 14 + 10 = 71.

Muse Spark Open 44

Read as software the owner runs. No account or key to run it. Meta's weights need manual approval on Hugging Face and an access token, and server authentication is the owner's to add (12 of 30). An agent model that calls whatever tools the operator gives it. The card recommends human-in-the-loop confirmation for irreversible actions and Meta reports agentic-risk evaluations, but nothing in the model or template enforces an approval (8 of 20). The chat template neutralises special-token sequences and media placeholders in caller text and wraps tool results in <tool_output>, and the card advises protecting internal assistant messages and layering defences against adversarial input. No indirect prompt-injection results are published (10 of 15). No call log of its own. Visibility depends on the serving stack (5 of 15). Weights are safetensors that load through Transformers' own classes from 5.19, and Meta published cyber, chemical, biological and misalignment evaluations. dev.meta.ai/.well-known/security.txt answers 404 (re-read on 4 November), research.meta.ai per the 15 October check, and no security route is on the card or in the docs, only museoss@meta.com and the Hugging Face discussions (9 of 20). 12 + 8 + 10 + 5 + 9 = 44.

Tencent Hy4 86

Model reading for hy4 on Tencent Cloud TokenHub. Platform facts are the tencent-tokenhub listing's (checked 1 November), carried over as RULES.md allows for a model on an already graded platform. Function calling is marked for hy4 in the model list (18 of 20). Structured output marked (13 of 15). Cache-hit price published, ¥0.3 against ¥6 input (14 of 15). 1M-token context with 960K input and 64K output (15). Batch at half price covered Hy3 and Hy4 preview on 1 November; whether it covers hy4 wasn't re-read (3 of 10). OpenAI and Anthropic SDKs with a changed base URL, plus Tencent Cloud's SDK and CLI for the control plane (8 of 10). Errors with code, messages, upstream status and request ID (15 of 15). 18 + 13 + 14 + 15 + 3 + 8 + 15 = 86.

Muse Spark Open 69

Model reading of the checklist (tool use, structured output, caching, context, throughput, defaults and SDKs, errors). OpenAI-format tools and parallel calls come back as tool_calls, and the template supports namespaced tools and deferred loading through a tool_search tool. A forced tool_choice is not documented (17 of 20). JSON mode and JSON Schema come from each runtime's guided decoding, which Meta's docs show for TensorRT-LLM only (8 of 15). Prefix caching is the runtime's. Meta's SGLang command turns on hierarchical caching, and Dynamo routes each request to the worker holding its cached context (8 of 10). 1,048,576-token context, with sliding-window attention on 46 of 62 layers holding KV memory down (15 of 15). Draft models for speculative decoding in SGLang and vLLM, and no batch mode of its own (7 of 10). Any OpenAI SDK works against a local server, but reasoning can't be turned off, defaults to the most expensive level, and needs streaming, a max_tokens budget for reasoning and the right stop tokens (6 of 15). The template raises named errors for an unsupported reasoning effort, tool arguments passed as a string and a tool message with no name. Server errors are the runtime's (8 of 15). 17 + 8 + 8 + 15 + 7 + 6 + 8 = 69.

Tencent Hy4 73

Model reading for hy4 on Tencent Cloud TokenHub. Platform facts are the tencent-tokenhub listing's (checked 1 November), carried over as RULES.md allows for a model on an already graded platform. Added on 4 November 2026 per the release notes; the model list and price page were updated on 9 November (30). Retirement notices of 21 to 33 days, no minimum written (7 of 12). Heavy retirement churn on the platform (3 of 8). Dated release notes and announcements, round-the-clock support (13 of 15). CLI replies not sampled (3 of 10). Control-plane SDKs and CLI, no inference SDK of its own (12 of 15). CLI runtime stated, CI not checked (5 of 10). 30 + 7 + 3 + 13 + 3 + 12 + 5 = 73. The model is a week old, so these lines are TokenHub's habits more than Hy4's record.

Muse Spark Open 60

Read for an open-weight model released three weeks ago. Released on 15 October 2026 (30). One release, BF16 and NVFP4 on the same day, with no history of revisions, successors or notice periods, and Meta Model API has no deprecation policy either (6 of 20). None of the 29 replies counted by script on 4 November across the BF16 repository's discussions came from Meta. Feedback goes to the discussions or museoss@meta.com, and Meta's docs and developer blog carry dated posts (8 of 25). Day-one support in vLLM, SGLang, TensorRT-LLM, Transformers, llama.cpp, MLX and Unsloth with Meta's own recipes, though SGLang's is a dev image and TensorRT-LLM's a feature branch (12 of 15). Current dependencies (Transformers 5.19), but no CI visible to us (4 of 10). 30 + 6 + 8 + 12 + 4 = 60.

Tencent Hy4 73

Model reading for hy4 on Tencent Cloud TokenHub. Platform facts are the tencent-tokenhub listing's (checked 1 November), carried over as RULES.md allows for a model on an already graded platform. No OpenAPI file for inference; field pages for the OpenAI and Anthropic formats as TokenHub implements them (12 of 25). llms.txt with Markdown copies (10). The model list, updated 9 November, gives hy4's context, input and output limits and marks thinking, structured output, function calling, caching and image understanding; computer use, which Tencent's release names, isn't marked (13 of 20). Typed field pages (10 of 15). Examples in five languages and an error page of 30 codes (14 of 15). Dated release notes, with hy4 added on 4 November (14 of 15). 12 + 10 + 13 + 10 + 14 + 14 = 73.

Muse Spark Open 61

Read for a model the owner serves. No specification of its own. The contract is the Jinja chat template, which renders OpenAI-format tools with additionalProperties: false added and rejects bad values, plus the OpenAI-compatible servers in vLLM and SGLang with the muse parsers (10 of 25). dev.meta.ai/llms.txt lists the Muse Spark Open overview and seven pages under it, plus three cookbook pages, each with a Markdown copy (10). The card lists seven intended uses and six limitations, the docs say when to pick BF16 or NVFP4 and which reasoning strength suits which task, and the Usage Policy states what is out of scope (17 of 20). reasoning_effort is an enum of four values and anything else returns an error, and tools take JSON Schema. Structured output is documented only for TensorRT-LLM's response_format, not in the prompting guide or for vLLM and SGLang (11 of 15). curl and Python examples for every runtime and hosted provider, the rendered prompt for each case, and the template's own error messages, but no error reference for a served model (11 of 15). One release, with no changelog or version scheme beyond the licence's version line, Muse Spark Open (2 of 15). 10 + 10 + 17 + 11 + 11 + 2 = 61.