Qwen 3.8 Max
Qwen 3.8 Max is a 2.4-trillion-parameter MoE model delivering a comprehensive leap in coding and professional work.
text + image → text · made by Alibaba
$1.65→$4.95per Mtok in / out
DeepInfra · 10 sellers
Sold by 16 ways
| Seller | Lane | Rate |
|---|---|---|
| DigitalOcean cloud | standard | $2per Mtok in$6per Mtok out$0.2per Mtok cachedmodels.dev · read 2026-10-31 |
| Nous Research 0 ways | $1.6per Mtok in$4.8per Mtok out$0.2per Mtok cached$2per Mtok cache write | |
| Nous Research aggregatornot on the seller's list today | standard | $1.6per Mtok in$4.8per Mtok out$0.2per Mtok cached$2per Mtok cache writeinference-api.nousresearch.com · read 2026-08-25 |
| DeepInfra aggregator | standard | $1.65per Mtok in$4.95per Mtok outapi.deepinfra.com · read 2026-08-25 |
| AtlasCloud aggregator | standard | $2per Mtok in$6per Mtok out$0.25per Mtok cachedapi.atlascloud.ai · read 2026-09-20 |
| ElectronHub aggregator | standard | $2per Mtok in$6per Mtok outapi.electronhub.ai · read 2026-10-31 |
| Fireworks AI 2 ways | $2per Mtok in$6per Mtok out$0.25per Mtok cached | |
| Fireworks AI aggregator | standard | $2per Mtok in$6per Mtok out$0.25per Mtok cachedmodels.dev · read 2026-10-31 |
| Fireworks AI aggregator | priority | $3per Mtok in$9per Mtok out$0.38per Mtok cachedraw.githubusercontent.com · read 2026-10-31 |
| Novita AI aggregator | standard | $2per Mtok in$6per Mtok out$0.25per Mtok cached$2.5per Mtok cache writeapi.novita.ai · read 2026-08-25 |
| OpenRouter 0 ways | $2per Mtok in$6per Mtok out$0.25per Mtok cached$2.5per Mtok cache write | |
| OpenRouter aggregatornot on the seller's list today | standard | $2per Mtok in$6per Mtok out$0.25per Mtok cached$2.5per Mtok cache writeopenrouter.ai · read 2026-08-24 |
| Requesty 3 ways | $2per Mtok in$6per Mtok out$0.25per Mtok cached | |
| Requesty aggregator | alibaba | $2per Mtok in$6per Mtok out$0.25per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
| Requesty aggregator | fireworks | $2per Mtok in$6per Mtok out$0.25per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
| Requesty aggregator | novita | $2per Mtok in$6per Mtok out$0.25per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
| Vercel AI Gateway aggregator | standard | $2per Mtok in$6per Mtok out$0.25per Mtok cached$2.5per Mtok cache writevercel.com · read 2026-08-25 |
| Vercel AI Gateway aggregatornot on the seller's list today | sandbox | $2per Mtok in$6per Mtok outvercel.com · read 2026-08-25 |
| Venice AI aggregator | standard | $2.5per Mtok in$7.5per Mtok out$0.31per Mtok cached$3.12per Mtok cache writemodels.dev · read 2026-10-31 |
| PPIO aggregator | standard | $12per Mtok in$36per Mtok outapi.ppinfra.com · read 2026-10-31 |
Measured 24 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 2ndof 165 | LMArena · Vision | Score (Elo) (max) | 1301 |
| 5thof 131 | Artificial Analysis · Agentic Index | Index | 56 |
| 10thof 141 | LMArena · WebDev | Score (Elo) (max) | 1672 |
| 11thof 135 | Webdev arena — Epoch AI | Arena Score | 1670.58 |
| 13thof 30 | Mirrorcode — Epoch AI | mean_score (xhigh) | 52.222 |
| 14thof 60 | Apex agents — Epoch AI | Pass@1 score | 0.633 |
| 16thof 163 | Artificial Analysis · Coding Index | Index | 76.2 |
| 18thof 306 | OTIS Mock AIME 2024-2025 | mean_score (xhigh) | 99.444 |
| 19thof 21 | Frontierswe — Epoch AI | Score (xhigh) | 15.809 |
| 19thof 142 | Mystery game puzzles — Epoch AI | mean_score (xhigh) | 38 |
| 22ndof 122 | Frontiermath tiers 1 3 v2 — Epoch AI | mean_score (xhigh) | 74.737 |
| 22ndof 414 | LMArena · Text | Score (Elo) (max) | 1483 |
| 23rdof 56 | Gdp pdf — Epoch AI | GDP.pdf score (xhigh) | 0.232 |
| 24thof 84 | Proofbench — Epoch AI | Accuracy | 58 |
| 27thof 154 | Chess puzzles — Epoch AI | mean_score (xhigh) | 46 |
| 27thof 326 | GPQA Diamond | mean_score (xhigh) | 92.677 |
| 31stof 78 | Frontiermath tier 4 v2 — Epoch AI | mean_score (xhigh) | 46.341 |
| 36thof 69 | Deepswe — Epoch AI | Pass@1 (xhigh) | 0.575 |
| 37thof 86 | Artificial Analysis · Intelligence Index | Artificial Analysis Intelligence Index | 34 |
| 40thof 93 | Simpleqa verified — Epoch AI | mean_score (xhigh) | 45.8 |
| 65thof 260 | Critpt — Epoch AI | Accuracy | 20 |
| 71stof 235 | Dtbench — Epoch AI | Accuracy (xhigh) | 92 |
| 85thof 197 | Lmca — Epoch AI | Score (xhigh) | 46.24 |
| 86thof 235 | Scicode — Epoch AI | Score | 53.241 |
About
Qwen 3.8 Max — a text model from Alibaba, sold by 10 companies from $1.65 in and $4.95 out per million tokens, placed 2nd of 165 on LMArena · Vision.
It takes text and images and returns text, with a context window of 1,000,000 tokens. It was published in July 2026. Its sellers say it can reason step by step and call a tool. The catalogue files it under reasoning, code and agents. Ten companies sell it. The cheapest is $1.65 in and $4.95 out per million tokens at DeepInfra. Beside the standard rate there are faster and separately routed lanes. It has been measured on 24 boards, and stands best at 2nd of 165 on LMArena · Vision.
Every current figure
- Maker
- Alibaba
- Register
- model
- Takes
- text + image
- Returns
- text
- Context
- 1,000,000 tokens
- Longest answer
- 131,072 tokens
- Published
- July 2026
- Licence
- not read
- Sellers
- 10
- Maker's own price
- not read
- Price
- $1.65 in and $4.95 out per million tokens — DeepInfra
- Boards
- 24
- Best place
- 2nd of 165 — LMArena · Vision
Known as 9 names
Qwen3.8 MaxQwen3.8 Max (xhigh)qwen3.8-max_xhighqwen3.8-maxqwen/qwen3.8-maxfireworks/qwen3p8-maxalibaba/qwen3.8-maxfireworks/qwen3.8-maxnovita/qwen/qwen3.8-max