DeepSeek-V3
DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token.
text → text · made by DeepSeek
Sold by 17 ways
| Seller | Lane | Rate |
|---|---|---|
| AWS Bedrock cloud | standard | $0.58per Mtok in$1.68per Mtok outraw.githubusercontent.com · read 2026-10-31 |
| Microsoft Azure AI cloud | standard | $1.14per Mtok in$4.56per Mtok outraw.githubusercontent.com · read 2026-10-31 |
| Hyperbolic aggregator | standard | $0.2per Mtok in$0.2per Mtok outraw.githubusercontent.com · read 2026-10-31 |
| Requesty 2 ways | $0.23per Mtok in$0.64per Mtok out | |
| Requesty aggregatornot on the seller's list today | deepinfra flex | $0.23per Mtok in$0.64per Mtok outrouter.requesty.ai · read 2026-08-25 |
| Requesty aggregator | deepinfra | $0.85per Mtok in$0.9per Mtok out$0.85per Mtok cachedrouter.requesty.ai · read 2026-10-09 |
| Requesty aggregator | novita | $0.89per Mtok in$0.89per Mtok out$0.89per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
| SiliconFlow aggregator | standard | $0.25per Mtok in$1per Mtok outmodels.dev · read 2026-10-31 |
| Vercel AI Gateway 0 ways | $0.27per Mtok in$1.12per Mtok out$0.14per Mtok cached | |
| Vercel AI Gateway aggregatornot on the seller's list today | standard | $0.27per Mtok in$1.12per Mtok out$0.14per Mtok cachedai-gateway.vercel.sh · read 2026-08-25 |
| DeepInfra aggregator | standard | $0.32per Mtok in$0.89per Mtok outapi.deepinfra.com · read 2026-08-25 |
| Hugging Face Inference Providers 2 ways | $0.4per Mtok in$1.3per Mtok out | |
| Hugging Face Inference Providers aggregator | deepinfra | $0.32per Mtok in$0.89per Mtok outrouter.huggingface.co · read 2026-08-25 |
| Hugging Face Inference Providers aggregator | standard | $0.4per Mtok in$1.3per Mtok outmodels.dev · read 2026-10-31 |
| Hugging Face Inference Providers aggregatornot on the seller's list today | novita | $0.4per Mtok in$1.3per Mtok outrouter.huggingface.co · read 2026-08-25 |
| ElectronHub 0 ways | $0.5per Mtok in$1.5per Mtok out | |
| ElectronHub aggregatornot on the seller's list today | standard | $0.5per Mtok in$1.5per Mtok outapi.electronhub.ai · read 2026-10-18 |
| Novita AI aggregator | standard | $0.89per Mtok in$0.89per Mtok outapi.novita.ai · read 2026-08-25 |
| Fireworks AI aggregator | standard | $0.9per Mtok in$0.9per Mtok outraw.githubusercontent.com · read 2026-10-31 |
| Together AI aggregator | standard | $1.25per Mtok in$1.25per Mtok outmodels.dev · read 2026-10-31 |
| PPIO 0 ways | $2per Mtok in$8per Mtok out | |
| PPIO aggregatornot on the seller's list today | standard | $2per Mtok in$8per Mtok outapi.ppinfra.com · read 2026-09-28 |
Measured 26 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 2ndof 134 | Arc ai2 — Epoch AI | Challenge score | 0.953 |
| 2ndof 88 | Bbh — Epoch AI | Average | 0.875 |
| 3rdof 217 | Mmlu — Epoch AI | EM | 0.872 |
| 5thof 113 | Hella swag — Epoch AI | Overall accuracy | 0.889 |
| 7thof 137 | Wino grande — Epoch AI | Accuracy | 85.2 |
| 18thof 54 | Live bench — Epoch AI | Global average | 60.45 |
| 18thof 115 | Trivia qa — Epoch AI | EM | 0.829 |
| 27thof 111 | Vectara Hallucination Leaderboard | Hallucination Rate | 6.1 |
| 32ndof 49 | Metr time horizons — Epoch AI | average_score | 0.474 |
| 41stof 72 | Aider polyglot — Epoch AI | Percent correct | 48.4 |
| 50thof 108 | Math level 5 — Epoch AI | mean_score | 64.851 |
| 59thof 92 | Forecastbench — Epoch AI | Overall score | 62.6 |
| 87thof 101 | FrontierMath (Tiers 1-3) | mean_score | 1.724 |
| 97thof 102 | Simplebench — Epoch AI | Score (AVG@5) | 0.189 |
| 110thof 150 | Design Arena · 3d (models) | Elo | 1084 |
| 113thof 114 | Design Arena · svg (models) | Elo | 973 |
| 127thof 157 | Design Arena · uicomponent (models) | Elo | 1088 |
| 128thof 154 | Design Arena · dataviz (models) | Elo | 1089 |
| 131stof 163 | Design Arena · codecategories (models) | Elo | 1106 |
| 131stof 169 | Design Arena · website (models) | Elo | 1118 |
| 137thof 162 | Design Arena · gamedev (models) | Elo | 1053 |
| 206thof 414 | LMArena · Text | Score (Elo) | 1358 |
| 207thof 235 | Scicode — Epoch AI | Score | 35.764 |
| 214thof 260 | Critpt — Epoch AI | Accuracy | 0 |
| 224thof 326 | GPQA Diamond | mean_score | 56.534 |
| 251stof 306 | OTIS Mock AIME 2024-2025 | mean_score | 15.833 |
About
DeepSeek-V3 — a text model from DeepSeek, sold by 10 companies from $0.2 in and $0.2 out per million tokens, placed 2nd of 134 on Arc ai2 — Epoch AI.
It takes text and returns text, with a context window of 163,840 tokens. It was published in December 2024, trained on material up to July 2024. Its sellers say it can call a tool. The catalogue files it under reasoning and agents. Ten companies sell it. The cheapest is $0.2 in and $0.2 out per million tokens at Hyperbolic. Beside the standard rate there is a separately routed lane. It has been measured on 26 boards, and stands best at 2nd of 134 on Arc ai2 — Epoch AI.
Every current figure
- Maker
- DeepSeek
- Register
- model
- Takes
- text
- Returns
- text
- Context
- 163,840 tokens
- Published
- December 2024
- Knowledge to
- July 2024
- Sellers
- 10
- Maker's own price
- not read
- Price
- $0.2 in and $0.2 out per million tokens — Hyperbolic
- Boards
- 26
- Best place
- 2nd of 134 — Arc ai2 — Epoch AI
Known as 8 names
DeepSeek-V3 (2025-03-24)V3deepseek-ai/DeepSeek-V3deepseek/deepseek_v3deepinfra/deepseek-ai/DeepSeek-V3deepinfra/deepseek-ai/DeepSeek-V3:flexnovita/deepseek/deepseek_v3deepseek/deepseek-v3