DeepSeek V4 Pro 0813
Sold by 21 sellers, timed at 63.4 tok/s, scored 23.1 on the Intelligence Index, and with results on 2 openly licensed boards.
retrieved sha256 · 0bb53a77713f
The same weights served by 3 providers, each at its own list price. OpenRouter's rate is its routed rate across whichever of them it uses, and it need not match any one of them.
What sellers charge
21 sellers · 19.82x from cheapest to dearestOne tick per seller, on a log scale of input price per million from $0.0999 at Relace to $1.98 at Parasail. 1 seller undercuts the $0.111 listed; 19 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 fp8 group alone spans 91% 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
104,214 requests · 21 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
21 priced · 21 timed · cheapest first- Relacerelace/fp4$0.0999$8.00$0.0999fp4393k62 tok/s26.3843ms1,916 req100%
- Waferwafer/us · listed$0.111$4.44$0.097mxfp4944knot matchednonenonenot matched99.93%
- Baidubaidu/fp8 · 90% off$0.132$0.396$0.0042fp8393k43 tok/s261.03s32,191 req99.58%
- Waferwafer/mxfp4$0.285$4.30$0.284mxfp4944knot matchednonenonenot matched99.86%
- Alibabaalibaba$0.5808$1.7424$0.05808not declared393knot matchednonenonenot matched100%
- Alibabaalibaba/us-east-1$0.66$1.98$0.066not declared393knot matchednonenonenot matched100%
- DeepSeekdeepseek$0.66$1.98$0.022not declared393k51 tok/s341.27s14,191 req100%
- StreamLakestreamlake · 50% off$0.66$1.98$0.022not declared384k35 tok/s211.34s5,360 req100%
- DeepInfradeepinfra/fp8/flex$1.04$2.08$0.08fp8393knot matchednonenonenot matched100%
- Novitanovita/fp8 · 5% off$1.254$3.762$0.0418fp8393k39 tok/s191.91s232 req100%
- DeepInfradeepinfra/fp8$1.30$2.60$0.10fp8393knot matchednonenonenot matched99.84%
- CoreWeavecoreweave/fp8$1.31$3.96$0.044fp8944k72 tok/s44444ms6,549 req100%
- BaseTenbaseten/fp4$1.32$3.96$0.132fp4262k96 tok/s48.6522ms351 req100%
- DigitalOceandigitalocean$1.32$3.96$0.044not declared944k92 tok/s60865ms13,450 req100%
- Nebiusnebius/fp8$1.32$3.96$0.04fp8944k67 tok/s20842ms277 req99.40%
- Parasailparasail/fp8$1.32$3.96$0.044fp8944knot matchednonenonenot matched100%
- SiliconFlowsiliconflow/fp8$1.32$3.96$0.044fp8393k33 tok/s24984ms117 req100%
- Togethertogether$1.32$3.96$0.13not declared944k93 tok/s52390ms5,896 req100%
- Venicevenice$1.65$4.95$0.165not declared262k40 tok/s191.78s26 req100%
- DeepInfradeepinfra/fp8/priority$1.95$3.90$0.15fp8393knot matchednonenonenot matched100%
- Parasailparasail/fast$1.98$5.94$0.066fp8944knot matchednonenonenot matched100%
- Alibabanot matchednot matchednot matchednot matchedno offer24 tok/s20968ms196 reqno offer
- Alibabanot matchednot matchednot matchednot matchedno offer33 tok/s201.01s1,553 reqno offer
- DeepInfranot matchednot matchednot matchednot matchedno offer46 tok/s251.86s17 reqno offer
- DeepInfranot matchednot matchednot matchednot matchedno offer57 tok/s271.21s207 reqno offer
- DeepInfranot matchednot matchednot matchednot matchedno offer64 tok/s371.3s443 reqno offer
- Parasailnot matchednot matchednot matchednot matchedno offer69 tok/s53244ms2,625 reqno offer
- Parasailnot matchednot matchednot matchednot matchedno offer65 tok/s49.1440ms66 reqno offer
- Wafernot matchednot matchednot matchednot matchedno offer77 tok/s49266ms5,618 reqno offer
- Wafernot matchednot matchednot matchednot matchedno offer89 tok/s52284ms12,933 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: 11 of the 21 timed routes are slower than it, and the middle route runs at 62 tok/s. The widest median-to-p90 gap is Nebius's, 3.35x.
Alibaba publishes 2 offers and had 2 routes timed; DeepInfra publishes 3 offers and had 3 routes timed; Parasail 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 . 19 of the 29 sellers with a reading fell under 99% on at least one of the 23 days held, 23 Oct 2026 to 21 Nov 2026: Cloudflare at 61.43% on 19 Nov 2026, DeepSeek at 62.43% on 2 Nov 2026, Venice at 91.55% on 6 Nov 2026, GMICloud at 92.42% on 7 Nov 2026, and 15 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 · 21 sellers- include_reasoningevery seller
- max_tokensevery seller
- reasoningevery seller
- reasoning_effortevery seller
- temperatureevery seller
- tool_choiceevery seller
- toolsevery seller
- top_pevery seller
- response_format20 of 21not BaseTen
- stop19 of 21not DigitalOcean and SiliconFlow
- frequency_penalty18 of 21not Baidu, DigitalOcean, and StreamLake
- presence_penalty17 of 21
- structured_outputs16 of 21
- top_k16 of 21
- repetition_penalty12 of 21
- seed12 of 21
- logit_bias10 of 21
- logprobs10 of 21
- top_logprobs10 of 21
- min_p7 of 21
8 of these 20 are accepted by every one of the 21 sellers that publish a list; the other 12 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 $1.193 per million tokens, 3 in to 1 out: cheaper than 133 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 boardEach 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.
Not ranked in any Design Arena category.
Position 43 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
2 of 6 boards · 31 results| Task | All runs | Score |
|---|---|---|
| Chess Puzzles | 167 | 0.46 |
| FrontierMath-Tier-4-v2-Private | 84 | 0.27 |
| FrontierMath-Tiers-1-3-v2-Private | 128 | 0.65 |
| GPQA diamond | 342 | 0.92 |
| MirrorCode | 48 | 0.33 |
| Mystery Game Puzzles | 149 | 0.43 |
| OTIS Mock AIME 2024-2025 | 319 | 0.99 |
| SimpleQA Verified | 106 | 0.53 |
Mean accuracy, 0 to 1, on the task the row names, as Epoch AI runs it and averaged over repeated runs. All eighteen of their tasks, including the held-out FrontierMath tiers, named as they name them. The model id carries the effort setting the run used. Epoch AI Benchmarking Hub names this run deepseek-v4-pro-0813_max. All runs is how many runs the publisher scored on the task, every model's. Read from epoch.ai ↗, which dates its own data . Epoch AI Benchmarking Hub data, by Epoch AI, under CC BY 4.0.
| Task | Score |
|---|---|
| AMPS_Hard | 98 |
| code_completion | 78.26 |
| code_generation | 76.06 |
| connections | 100 |
| consecutive_events | 90.23 |
| integrals_with_game | 94 |
| javascript | 68.18 |
| logic_with_navigation | 66 |
| math_comp | 97.06 |
| olympiad | 91.29 |
| paraphrase | 68.63 |
| plot_unscrambling | 64.22 |
| python | 50 |
| simplify | 61.57 |
| spatial | 96 |
| story_generation | 74.93 |
| summarize | 65.67 |
| tablejoin | 49.46 |
| tablereformat | 98.04 |
| theory_of_mind | 84.61 |
| typescript | 46.67 |
| typos | 82 |
| zebra_puzzle | 96.75 |
Percent correct on the task the row names, higher is better. All 23 published columns; there is no average to carry, because the site computes its headline in the browser from a weighting this does not reimplement. LiveBench names this run deepseek-v4-pro-0813, with no effort setting written on it. Read from livebench.ai ↗, which dates its own data . LiveBench, Apache-2.0. Results produced by the LiveBench maintainers on questions refreshed monthly.
2 of the 6 openly licensed boards Vitna mirrors carry a result for DeepSeek V4 Pro 0813, each printed as its publisher published it and none added to another. MTEB, HELM, Vectara hallucination leaderboard and Aider polyglot have not published one under a name that matches its listing, and names are matched to the catalog exactly and never guessed. A task that is a link opens its column on /capability, with every run scored on it; an effort that is a link opens its run's row there, where the run is ranked on Vitna's scale. 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.
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.
