Qwen3 8B
Qwen3-8B is a dense 8.2B parameter causal language model from the Qwen3 series, designed for both reasoning-heavy tasks and efficient dialogue.
text → text · made by Alibaba
Sold by 8 ways
| Seller | Lane | Rate |
|---|---|---|
| Thinking Machines Lab api | training | $0.2per Mtok in$0.6per Mtok out$0.039per Mtok cachedtinker-docs.thinkingmachines.ai · read 2026-08-25 |
| Novita AI aggregator | standard | $0.035per Mtok in$0.14per Mtok outapi.novita.ai · read 2026-08-25 |
| SiliconFlow aggregator | standard | $0.06per Mtok in$0.06per Mtok outmodels.dev · read 2026-09-16 |
| Hugging Face Inference Providers aggregator | nscale | $0.07per Mtok in$0.18per Mtok outrouter.huggingface.co · read 2026-08-25 |
| Inference.net aggregator | standard | $0.12per Mtok in$0.46per Mtok outapi.inference.net · read 2026-09-16 |
| Nous Research aggregator | standard | $0.12per Mtok in$0.46per Mtok outinference-api.nousresearch.com · read 2026-09-12 |
| OpenRouter aggregator | standard | $0.12per Mtok in$0.46per Mtok outopenrouter.ai · read 2026-08-24 |
| Fireworks AI aggregator | standard | $0.2per Mtok in$0.2per Mtok outraw.githubusercontent.com · read 2026-09-16 |
Measured 11 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 9thof 105 | Vectara Hallucination Leaderboard | Hallucination Rate | 4.8 |
| 114thof 123 | Lmca — Epoch AI | Score | 8.82 |
| 120thof 161 | Dtbench — Epoch AI | Accuracy | 59.73 |
| 136thof 143 | Artificial Analysis · Agentic Index | Index | 0.8 |
| 158thof 222 | Chess puzzles — Epoch AI | mean_score | 5 |
| 162ndof 168 | Scicode — Epoch AI | Score | 22.569 |
| 168thof 290 | OTIS Mock AIME 2024-2025 | mean_score | 56.111 |
| 175thof 177 | Critpt — Epoch AI | Accuracy | 0 |
| 179thof 184 | Artificial Analysis · Coding Index | Index | 9 |
| 210thof 313 | GPQA Diamond | mean_score | 56.755 |
| 487thof 656 | Epoch capabilities index — Epoch AI | ECI Score (none) | 136.19 |
About
Qwen3 8B — a text model from Alibaba, sold by 8 companies from $0.035 in and $0.14 out per million tokens, placed 9th of 105 on Vectara Hallucination Leaderboard.
It takes text and returns text, with a context window of 131,072 tokens. It was published in 2024. Its sellers say it can reason step by step and call a tool. The catalogue files it under reasoning. Eight companies sell it. The cheapest is $0.035 in and $0.14 out per million tokens at Novita AI. Beside the standard rate there is a separately routed lane. It has been measured on 11 boards, and stands best at 9th of 105 on Vectara Hallucination Leaderboard.
Every current figure
- Maker
- Alibaba
- Register
- model
- Takes
- text
- Returns
- text
- Context
- 131,072 tokens
- Longest answer
- 131,000 tokens
- Published
- 2024
- Parameters
- 8.2 billion
- Licence
- not read
- Sellers
- 8
- Maker's own price
- not read
- Price
- $0.035 in and $0.14 out per million tokens — Novita AI
- Boards
- 11
- Best place
- 9th of 105 — Vectara Hallucination Leaderboard
Known as 2 names
Qwen/Qwen3-8Bqwen/qwen3-8b-fp8