Hy4 preview
Sold by 5 sellers, timed at 63.8 tok/s, and not scored by Artificial Analysis.
retrieved sha256 · 0bb53a77713f
The same weights, served by one provider at its own list price, read from novita.ai. OpenRouter's rate is its routed rate across whichever sellers it uses, and the two need not agree.
What sellers charge
5 sellers · 1.25x from cheapest to dearestOne tick per seller, on a log scale of input price per million from $0.834 at DeepInfra to $1.0425 at Venice. The $0.7506 listed is the cheapest on offer; 5 sellers are dearer. A faint tick declares no number format. Every seller is named on the endpoint ledger below.
The same sellers grouped by the number format each declares, on the same scale. The fp4 group alone spans 68% of the range, so the cheap end is not cheap for its number format. A declaration is checked by nobody: this groups the rows and certifies none.
Measured speed
876 requests · 5 endpoints · 30-minute windowOpenRouter's own timing of its routed traffic, weighted by requests, over one 30-minute window read . Every route is on the endpoint ledger below.
Every endpoint
5 priced · 5 timed · cheapest first- DeepInfradeepinfra/fp4$0.834$2.501$0.042fp4131k59 tok/s34496ms84 req100%
- Novitanovita/fp8$0.834$2.501$0.042fp864k47 tok/s214.21s126 req100%
- SiliconFlowsiliconflow/fp8$0.834$2.501$0.042fp8262k66 tok/s42.41.2s281 req100%
- Tencenttencent/fp8$0.834$2.501$0.042fp864k61 tok/s372.52s218 req100%
- Venicevenice/fp4$1.0425$3.12625$0.0525fp464k79 tok/s52.61.6s167 req99.43%
Each track runs from a route's p90, the rate nine in ten of its requests met or beat, up to its median. Weighted by requests, the figure above reports where the traffic went rather than what one route does: 3 of the 5 timed routes are slower than it, and the middle route runs at 61 tok/s. The widest median-to-p90 gap is Novita's, 2.24x.
Uptime is OpenRouter's own figure for each seller over the trailing day, read once a day, and the column prints the latest reading, taken . 3 of the 9 sellers with a reading fell under 99% on at least one of the 23 days held, 23 Oct 2026 to 21 Nov 2026: GMICloud at 80.00% on 2 Nov 2026, Cohere (cohere/mxfp4) at 97.23% on 7 Nov 2026, and Cohere (cohere) at 97.57% on 4 Nov 2026. Source: OpenRouter's public endpoints API, one reading a day for each seller: 1,211 sellers carried one on the latest of the 23 days held.
Source: OpenRouter's public endpoints API. 1257 sellers across 359 models when the mirror ran, 186 of them served by more than one. List prices, not what anybody paid; a discount is the seller's own. Read . Speed is OpenRouter's own timing over one 30-minute window, read .
What each seller accepts
19 parameters · 5 sellers- include_reasoningevery seller
- max_tokensevery seller
- reasoningevery seller
- reasoning_effortevery seller
- structured_outputsevery seller
- temperatureevery seller
- tool_choiceevery seller
- toolsevery seller
- frequency_penalty4 of 5not Tencent
- stop4 of 5not SiliconFlow
- top_k4 of 5not Tencent
- top_p4 of 5not Tencent
- presence_penalty3 of 5not SiliconFlow and Tencent
- response_format3 of 5not Novita and Venice
- repetition_penalty2 of 5not SiliconFlow, Tencent, and Venice
- seed2 of 5not SiliconFlow, Tencent, and Venice
- logit_bias1 of 5only DeepInfra
- max_completion_tokens1 of 5only Tencent
- min_p1 of 5only DeepInfra
8 of these 19 are accepted by every one of the 5 sellers that publish a list; the other 11 are declared by some sellers and not others, so what a request sending one of them gets depends on which seller serves it. Each list is the seller's own declaration in OpenRouter's endpoints API, read , and nothing here tested one.
Price position
342 priced modelsBlended $1.126 per million tokens, 3 in to 1 out: cheaper than 135 of the 342 priced models in the catalog. Each bar is one stretch of the catalog on a logarithmic price scale; the mark is this listing.
Measured quality
145 models with an index · 132 on a design arena boardNot ranked in any Design Arena category.
Epoch AI has not placed Hy4 preview on its Capabilities Index, or places it under a name the pairing does not reach. Epoch places a model once four of its benchmarks carry a result.
Benchmark results
none of 6 boardsNone of the 6 openly licensed boards Vitna mirrors has published a result for Hy4 preview under a name that matches its listing, and names are matched to the catalog exactly and never guessed. Every board, with its licence and what its publisher measures, is on /benchmarks.
Source: OpenRouter (openrouter.ai/models), as of 17 Nov 16:29 UTC. sha256 0bb53a77713f over 853,270 bytes.
Catalog: OpenRouter's public model list.
Intelligence, coding and agentic indexes: Artificial Analysis, as published by OpenRouter.
Elo, win rate and rank: Design Arena, as published by OpenRouter.
Source: each model's own Hugging Face repository (huggingface.co), as of 20 Nov 2026. 162 repositories read, 4 did not answer.
Licence, parameter count, model class and gating: each model's own Hugging Face repository, the one OpenRouter names for the listing.
A licence is the tag the uploader put on the repository, read from nothing else and verified by nobody.
A parameter count is the total the repository's safetensors files declare; for a mixture of experts that is every expert, not the share one token runs through.
Vitna ran none of these evaluations and adjusted none of these figures.
