Get started

Kimi K3.1 Pro (batch)

Sold by 28 sellers, timed at 82.7 tok/s, scored 46.5 on the Intelligence Index, and on 16 Design Arena boards.

retrieved sha256 · 0bb53a77713f

Models/MoonshotAI/Kimi K3.1 Pro (batch)USD per 1M tokens · as OpenRouter routes it
moonshotai/kimi-k3.1-pro:batchMoonshotAI · batch variant · Other tokenizer
OpenRouter list price · the routed rate, as the catalog carried itretrieved 17 Nov 16:29 UTC from openrouter.ai ↗
Input$0.372per 1M tokens
Output$5.583per 1M tokens
Cached input$0.10per 1M tokens
Context1.05M1,048,576 tokens
What the list says · the fields OpenRouter publishes for this id12 fields
Listed
Canonical idmoonshotai/kimi-k3.1-pro-20261023
Context1,048,576 tokens
Max output943,718 tokens
Modalitiestext, image, video → text
TokenizerOther
Reasoningoptional, on by default · effort max, high, low, none, max by default
Moderationnot moderated by OpenRouter
Variantbatch
Other listingbase listing
What the weights say · read from the model's own repository on Hugging Faceretrieved from huggingface.co ↗
Licencekimi-k3.1-pro, tagged other
Parameters2.8T · 2,780,110,363,392 declared
Model classKimiK31ForConditionalGeneration
Downloadnot gated
Published

What sellers charge

28 sellers · 18.31x from cheapest to dearest

One tick per seller, on a log scale of input price per million from $0.2403 at Relace to $4.40 at Amazon Bedrock. 2 sellers undercut the $0.372 listed; 26 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 mxfp4 group alone spans 95% 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

141,675 requests · 28 endpoints · 30-minute window
Throughput82.7 tok/sMedian output rate, weighted across the 28 endpoints serving it by the requests each answered.the speed board
First token1.07sMedian wait before the first token, over the same window and weighted the same way.how Vitna reads a figure
Endpoint spread27 to 118The 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

28 priced · 28 timed · cheapest first
  1. Relacerelace/fp4$0.2403$8.4023$0.20fp4944k27 tok/s121.38s13,055 req99.93%
  2. Waferwafer/us$0.253$8.80$0.18mxfp4944knot matchednonenonenot matched99.93%
  3. Sail Researchsail-research/fp4/us$1.0425$9.35$0.187fp4944knot matchednonenonenot matched100%
  4. Decartdecart/mxfp4 · 40% off$1.20$6.00$0.12mxfp4944k59 tok/s372.1s3,895 req100%
  5. Morphmorph/fp8$1.375$9.00$0.172fp8944k62 tok/s381.32s2,751 req100%
  6. Sail Researchsail-research/fp4$1.4025$9.35$0.187fp4944knot matchednonenonenot matched100%
  7. AkashMLakashml/fp4$1.75$10.00$0.18fp4944k63 tok/s301.85s568 req100%
  8. InferenceNetinference-net/mxfp4$1.80$9.00$0.198mxfp4944knot matchednonenonenot matched100%
  9. InferenceNetinference-net/mxfp4/us$1.80$9.00$0.198mxfp4944knot matchednonenonenot matched100%
  10. Waferwafer/mxfp4$1.80$9.00$0.18mxfp4944knot matchednonenonenot matched99.86%
  11. DeepInfradeepinfra/bf16 · 5% off$1.90$9.50$0.19bf16944k57 tok/s431.24s1,003 req98.26%
  12. DigitalOceandigitalocean$1.90$9.50$0.19not declared944k85 tok/s411.46s8,491 req100%
  13. Alibabaalibaba$2.00$10.00$0.20not declared944knot matchednonenonenot matched100%
  14. Fireworksfireworks$2.00$10.00$0.20not declared944k45 tok/s291.45s3,633 req98.05%
  15. Makoramakora$2.00$9.49$0.20not declared944k66 tok/s41827ms833 req99.55%
  16. Modalmodal/mxfp4$2.00$10.00$0.20mxfp4944k106 tok/s61643ms15,153 req100%
  17. Moonshot AImoonshotai/mxfp4$2.00$10.00$0.20mxfp4944knot matchednonenonenot matched100%
  18. Nebiusnebius/fp4$2.00$10.00$0.20fp4944k110 tok/s381.33s14,908 req100%
  19. Parasailparasail/fp4$2.00$10.00$0.20fp4944k101 tok/s471.2s5,435 req100%
  20. PrimaLabsprimalabs/mxfp4$2.00$10.00$0.20mxfp4944k86 tok/s441.14s1,873 req99.86%
  21. Riverriver$2.00$10.00$0.20not declared1.05M81 tok/s57641ms4,552 req100%
  22. Togethertogether$2.00$10.00$0.20not declared944k40 tok/s251.49s1,746 req99.93%
  23. InferenceNetinference-net/fast$2.70$13.50$0.27mxfp4944knot matchednonenonenot matched100%
  24. InferenceNetinference-net/us/fast$2.70$13.50$0.27mxfp4944knot matchednonenonenot matched100%
  25. Alibabaalibaba/prime$4.00$20.00$0.20not declared944knot matchednonenonenot matched100%
  26. Amazon Bedrockamazon-bedrock/us-east-2$4.00$20.00$0.20not declared128knot matchednonenonenot matched96.51%
  27. Moonshot AImoonshotai/highspeed$4.00$20.00$0.20mxfp4944knot matchednonenonenot matched100%
  28. Amazon Bedrockamazon-bedrock/us-east-1$4.40$22.00$0.22not declared128knot matchednonenonenot matched98.14%
  29. Alibabanot matchednot matchednot matchednot matchedno offer45.5 tok/s223.14s221 reqno offer
  30. Alibabanot matchednot matchednot matchednot matchedno offer35 tok/s242.09s162 reqno offer
  31. Amazon Bedrocknot matchednot matchednot matchednot matchedno offer56 tok/s195.57s340 reqno offer
  32. Amazon Bedrocknot matchednot matchednot matchednot matchedno offer98 tok/s37.71.96s153 reqno offer
  33. InferenceNetnot matchednot matchednot matchednot matchedno offer107 tok/s60741ms3,624 reqno offer
  34. InferenceNetnot matchednot matchednot matchednot matchedno offer114 tok/s68547ms10,283 reqno offer
  35. InferenceNetnot matchednot matchednot matchednot matchedno offer118 tok/s73529ms11,347 reqno offer
  36. InferenceNetnot matchednot matchednot matchednot matchedno offer114 tok/s69694ms4,300 reqno offer
  37. Moonshot AInot matchednot matchednot matchednot matchedno offer29 tok/s213.04s3,815 reqno offer
  38. Moonshot AInot matchednot matchednot matchednot matchedno offer47 tok/s25.63.82s60 reqno offer
  39. Sail Researchnot matchednot matchednot matchednot matchedno offer55 tok/s312.01s1,152 reqno offer
  40. Sail Researchnot matchednot matchednot matchednot matchedno offer66 tok/s321.75s809 reqno offer
  41. Wafernot matchednot matchednot matchednot matchedno offer68 tok/s26801ms21,159 reqno offer
  42. Wafernot matchednot matchednot matchednot matchedno offer68 tok/s28635ms6,354 reqno offer

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: 18 of the 28 timed routes are slower than it, and the middle route runs at 66 tok/s. The widest median-to-p90 gap is Amazon Bedrock's, 2.95x.

Alibaba publishes 2 offers and had 2 routes timed; Amazon Bedrock publishes 2 offers and had 2 routes timed; InferenceNet publishes 4 offers and had 4 routes timed; Moonshot AI publishes 2 offers and had 2 routes timed; Sail Research publishes 2 offers and had 2 routes timed; Wafer publishes 2 offers and had 2 routes timed. The two documents share no route id, so those offers and routes are printed apart rather than paired.

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 . 16 of the 30 sellers with a reading fell under 99% on at least one of the 23 days held, 23 Oct 2026 to 21 Nov 2026: Amazon Bedrock (amazon-bedrock/us-east-2) at 65.04% on 18 Nov 2026, Amazon Bedrock (amazon-bedrock/us-east-1) at 70.34% on 18 Nov 2026, Together at 77.47% on 16 Nov 2026, DeepInfra at 87.50% on 6 Nov 2026, and 12 more. 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

20 parameters · 28 sellers
  1. include_reasoningevery seller
  2. max_tokensevery seller
  3. reasoningevery seller
  4. reasoning_effortevery seller
  5. toolsevery seller
  6. response_format26 of 28not Amazon Bedrock (amazon-bedrock/us-east-1) and Amazon Bedrock (amazon-bedrock/us-east-2)
  7. temperature26 of 28not Amazon Bedrock (amazon-bedrock/us-east-1) and Amazon Bedrock (amazon-bedrock/us-east-2)
  8. tool_choice26 of 28not Amazon Bedrock (amazon-bedrock/us-east-1) and Amazon Bedrock (amazon-bedrock/us-east-2)
  9. top_p26 of 28not Amazon Bedrock (amazon-bedrock/us-east-1) and Amazon Bedrock (amazon-bedrock/us-east-2)
  10. stop25 of 28not Alibaba (alibaba), Alibaba (alibaba/prime), and DigitalOcean
  11. structured_outputs25 of 28not Amazon Bedrock (amazon-bedrock/us-east-1), Amazon Bedrock (amazon-bedrock/us-east-2), and Relace
  12. presence_penalty23 of 28
  13. frequency_penalty21 of 28
  14. seed21 of 28
  15. top_k20 of 28
  16. logprobs18 of 28
  17. top_logprobs18 of 28
  18. repetition_penalty15 of 28
  19. logit_bias12 of 28
  20. min_p10 of 28

5 of these 20 are accepted by every one of the 28 sellers that publish a list; the other 15 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 $1.675 per million tokens, 3 in to 1 out: cheaper than 114 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 Index46.5Position 11 of 58, above 81% of the models scored.the intelligence leaderboard
Coding Indexnot scoredArtificial Analysis has not scored this model on its Coding Index.the coding leaderboard
Agentic Indexnot scoredArtificial Analysis has not scored this model on its Agentic Index.the agentic leaderboard
  1. Android native · Agents1157 Elowin rate 43.8% · source rank 25the board, 41 ranked
  2. codetovideo · Agents1255 Elowin rate 59.9% · source rank 8the board, 26 ranked
  3. Full stack · Agents1321 Elowin rate 61.8% · source rank 3the board, 44 ranked
  4. Mobile apps · Agents1267 Elowin rate 59.5% · source rank 5the board, 49 ranked
  5. pdf · Agents1222 Elowin rate 55.5% · source rank 14the board, 36 ranked
  6. python-pptx slides · Agents1286 Elowin rate 60.3% · source rank 3the board, 29 ranked
  7. spreadsheet · Agents1318 Elowin rate 62.6% · source rank 4the board, 32 ranked
  8. Web apps · Agents1307 Elowin rate 60.0% · source rank 3the board, 47 ranked
  9. 3D scenes · Models1440 Elowin rate 70.1% · source rank 3the board, 113 ranked
  10. ASCII art · Models1296 Elowin rate 58.5% · source rank 8the board, 67 ranked
  11. Code · Models1348 Elowin rate 60.4% · source rank 7the board, 120 ranked
  12. Data visualisation · Models1293 Elowin rate 55.0% · source rank 13the board, 110 ranked
  13. Game development · Models1402 Elowin rate 65.6% · source rank 4the board, 120 ranked
  14. SVG · Models1324 Elowin rate 61.9% · source rank 4the board, 83 ranked
  15. UI components · Models1390 Elowin rate 65.5% · source rank 1the board, 116 ranked
  16. Websites · Models1319 Elowin rate 57.5% · source rank 7the board, 126 ranked
Epoch Capabilities Index · a fit by a third party over every benchmark this model has results onthe whole index, with intervals
Index161.2158.9 to 163.7
Position18 of 258above 93% of the models placed
Results behind it16benchmark results
As Epoch names itKimi K3.1 Proreleased 23 Oct 2026

Position 18 of the 258 models Epoch has placed, which is not the population of either board above, on a scale anchored at Claude 3.5 Sonnet 130 and GPT-5 150. Epoch states that absolute values are meaningless by themselves and that comparisons between models are what the index is for, that values are not linearly related to benchmark accuracy, and that a highly specialised model may score low while being capable in its own domain.

Benchmark results

none of 6 boards

None of the 6 openly licensed boards Vitna mirrors has published a result for Kimi K3.1 Pro (batch) 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.
Source: the Epoch AI Capabilities Index (epoch.ai/eci), as of 20 Nov 2026. 258 models placed over 2917 benchmark results.
Epoch AI Benchmarking Hub data, by Epoch AI, under CC BY 4.0.
Vitna ran none of these evaluations and adjusted none of these figures.