deepseek-reasoner
An alias for DeepSeek v4 Flash in thinking mode.
text → text · made by DeepSeek
$0.14→$0.28per Mtok in / out
Requesty
Sold by 1 way
| Seller | Lane | Rate |
|---|---|---|
| Requesty aggregator | deepseek | $0.14per Mtok in$0.28per Mtok out$0.028per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
Measured 5 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 14thof 72 | Aider polyglot — Epoch AI | Percent correct | 74.2 |
| 71stof 290 | OTIS Mock AIME 2024-2025 | mean_score | 87.817 |
| 97thof 313 | GPQA Diamond | mean_score | 83.424 |
| 114thof 222 | Chess puzzles — Epoch AI | mean_score | 14 |
| 299thof 656 | Epoch capabilities index — Epoch AI | ECI Score | 146.21 |
About
deepseek-reasoner — a text model from DeepSeek, placed 14th of 72 on Aider polyglot — Epoch AI.
It takes text and returns text, with a context window of 1,000,000 tokens. It was published in December 2025, trained on material up to September 2025. Its sellers say it can reason step by step and call a tool. The catalogue files it under reasoning. It has been measured on 5 boards, and stands best at 14th of 72 on Aider polyglot — Epoch AI.
Every current figure
- Maker
- DeepSeek
- Register
- model
- Takes
- text
- Returns
- text
- Context
- 1,000,000 tokens
- Longest answer
- 384,000 tokens
- Published
- December 2025
- Knowledge to
- September 2025
- Boards
- 5
- Best place
- 14th of 72 — Aider polyglot — Epoch AI
Known as 1 name
deepseek/deepseek-reasoner