Get started

Qwen3.8 2.4T A95B

Timed at 144 tok/s and scored 71.9 on the Coding Index.

retrieved sha256 · 0bb53a77713f

Models/Qwen/Qwen3.8 2.4T A95BUSD per 1M tokens · as OpenRouter routes it
qwen/qwen3.8-2.4t-a95bQwen · base listing · Qwen tokenizer
OpenRouter list price · the routed rate, as the catalog carried itretrieved 17 Nov 16:29 UTC from openrouter.ai ↗
Input$2.00per 1M tokens
Output$6.00per 1M tokens
Cached input$0.25per 1M tokens
Context1.05M1,048,576 tokens
What Vitna read · the provider's own pricing page, kept as a dated recordin / out per 1M tokens
Novita AIopen weights$2.00 / $6.00confirmed 22 Nov 2026

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 the list says · the fields OpenRouter publishes for this id10 fields
AuthorQwen
Listed
Canonical idqwen/qwen3.8-2.4t-a95b-20260812
Context1,048,576 tokens
Max output131,072 tokens
Modalitiestext → text
TokenizerQwen
Reasoningalways on · effort xhigh, medium, low, xhigh by default
Moderationnot moderated by OpenRouter
What the weights say · read from the model's own repository on Hugging Faceretrieved from huggingface.co ↗
Parameters2.4T · 2,446,182,725,504 declared
Model classQwen3_5MoeForCausalLM
Downloadnot gated
Published

Measured speed

2,197 requests · 4 endpoints · 30-minute window
Throughput144 tok/sMedian output rate, weighted across the 4 endpoints serving it by the requests each answered.the speed board
First token1.39sMedian wait before the first token, over the same window and weighted the same way.how Vitna reads a figure
Endpoint spread34 to 148The slowest and the fastest route serving this model, in tokens per second. Speed is a property of the route, not of the weights.the board, fastest first

OpenRouter'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

4 priced · 4 timed · cheapest first
  1. Alibabaalibaba · listed$2.00$6.00$0.25not declared131k34 tok/s191.67s24 req100%
  2. Novitanovita · listed$2.00$6.00$0.25not declared131k39 tok/s225.58s41 req100%
  3. SiliconFlowsiliconflow/fp8 · listed$2.00$6.00$0.25fp8131k42 tok/s252.25s9 req100%
  4. Venicevenice · listed · 20% off$2.00$6.00$0.25not declared66k148 tok/s621.3s2,123 req99.83%

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 4 timed routes are slower than it, and the middle route runs at 40.5 tok/s. The widest median-to-p90 gap is Venice's, 2.39x.

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 . 2 of the 5 sellers with a reading fell under 99% on at least one of the 23 days held, 23 Oct 2026 to 21 Nov 2026: Venice at 87.86% on 9 Nov 2026 and SiliconFlow at 96.04% on 20 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

18 parameters · 4 sellers
  1. frequency_penaltyevery seller
  2. include_reasoningevery seller
  3. max_tokensevery seller
  4. reasoningevery seller
  5. reasoning_effortevery seller
  6. temperatureevery seller
  7. tool_choiceevery seller
  8. toolsevery seller
  9. top_kevery seller
  10. top_pevery seller
  11. presence_penalty3 of 4not SiliconFlow
  12. response_format3 of 4not Venice
  13. stop3 of 4not SiliconFlow
  14. structured_outputs3 of 4not Venice
  15. logprobs2 of 4not SiliconFlow and Venice
  16. seed2 of 4not SiliconFlow and Venice
  17. top_logprobs2 of 4not SiliconFlow and Venice
  18. repetition_penalty1 of 4not Alibaba, SiliconFlow, and Venice

10 of these 18 are accepted by every one of the 4 sellers that publish a list; the other 8 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 models

Blended $3.00 per million tokens, 3 in to 1 out: cheaper than 81 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 board

Each mark is one board this model stands on, placed at the share of that board it scores above, and led to its number in the list under the rule. The columns and the ledger below carry the figures.

Intelligence Indexnot scoredArtificial Analysis has not scored this model on its Intelligence Index.the intelligence leaderboard
Coding Index71.9Position 19 of 118, above 84% of the models scored.the coding leaderboard
Agentic Index50.1Position 10 of 94, above 89% of the models scored.the agentic leaderboard

Not ranked in any Design Arena category.

Epoch Capabilities Index · a fit by a third party over every benchmark this model has results onthe whole index, with intervals

Epoch AI has not placed Qwen3.8 2.4T A95B 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 boards

None of the 6 openly licensed boards Vitna mirrors has published a result for Qwen3.8 2.4T A95B 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.

Sources

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.