Page last updated on
Qwen3.8-Max
By Alibaba
The largest model in the Qwen family, a 2.4T-parameter MoE with 95B active parameters and a 1M-token context window. Built for long-horizon autonomous coding, professional work, and multimodal agents. Open weights are promised but not published yet.
Specifications
- Total parameters
- 2.4T (95B active)
- Active parameters
- 95B
- Architecture
- MoE
- Context window
- 1M
- Modality
- Text, Image, Video in, Text out
- Released
- August 2026
- Best for
- Long-horizon autonomous coding, professional work agents, and multimodal agents
Good to know
-
The weights are not public yet. Alibaba announced Qwen3.8-Max on August 3, 2026 as the first Qwen model at Max scale that will be open-sourced, and said the weights land on Hugging Face and ModelScope next week.
-
The API exposes a reasoning_effort parameter with three levels: xhigh (the default) for thorough analysis, medium for a balance of accuracy and speed, and low for cost-sensitive work. The preserve_thinking flag is on by default. QwenCloud serves chat completions, an OpenAI-compatible Responses API, and an Anthropic-compatible endpoint, so the model drops into Claude Code, Codex, OpenClaw, Qoder, and Qwen Code.
-
A companion open-weight model, Qwen3.8-27B, was announced for the same weight drop.
Architecture
Alibaba published parameter counts and very little else. Qwen3.8-Max is a Mixture-of-Experts model with 2.4 trillion total parameters and 95 billion active per token, built on the architectural foundation of Qwen3.5. The expert count, attention design, vocabulary size, and training precision are all unreported. That is unusual for a model billed as the first open-weight release at Max scale, and it should be filled in when the checkpoint lands.
What Alibaba did detail is the training system, and the argument there is that reinforcement learning environments matter more than architecture at this scale. Three pieces carry it:
- Decoupled environments that scale along independent axes: task, workspace, and harness. Growth compounds combinatorially instead of needing bespoke integration per environment.
- A universal reward system covering execution checks, rubric-based judging over text and rendered visuals, and agentic inspection, all under one reward model rather than per-task verifiers.
- An online data balancer that shapes every batch across tasks, difficulty, workspaces, and harnesses. That suppresses gradient variance between batches and keeps RL compute scaling stable.
The visible payoff is harness independence. Alibaba reports Qwen3.8-Max scoring within a few points of itself across QwenWork, Claude Code, Codex, OpenClaw, and Hermes on CoWorkBench, WorkspaceBench, and JobBench. Most models are tuned to one scaffold and drop elsewhere. On the other side of the ledger, the model spends a lot of tokens to get there. At the default xhigh reasoning effort, one independent tester measured a coding suite at $126.25 on Qwen3.8-Max against $13.60 on DeepSeek V4-Flash, roughly 9x, and called the model slow. So quantization and the lower effort levels will matter once weights arrive.
For more models like this one, browse the full open source LLM directory, read the open source LLM ecosystem statistics, or see how a 1M-token window is served in what is a KV cache.
Benchmarks
Alibaba published a full benchmark set against Claude Opus 4.8, Claude Fable 5, and GPT-5.6 Sol. Qwen3.8-Max leads the whole comparison group on PaperBench (93.0), IFBench (82.8), WideSearch (81.9), HealthBench (60.2), PLawBench (73.2), and PRBench-Finance (58.3). The weak spots are the hardest coding and reasoning tests: DeepSWE 1.1 (56.6), MLS-Bench-Lite (41.0), and HLE (43.6) all sit below the closed frontier.
Qwen3.8-Max vs frontier models
Scores published by Alibaba in the Qwen3.8-Max announcement. Qwen3.8-Max runs at xhigh reasoning effort. Harnesses differ by benchmark: most Qwen numbers come from the Claude Code harness, and competitor cells are often best published scores from other harnesses or leaderboards.
Each model keeps the same color across every benchmark, stacked top to bottom in this order. Hover or focus a row to read all six scores.
Separate axis. Elo is not a percentage, so this row is plotted as points on a fitted scale starting at 1,400.
Separate axis. Elo is not a percentage, so this row is plotted as points on a fitted scale starting at 1,400.
- Opus 4.8
- Fable 5
- GPT-5.6 Sol max
- Qwen3.7-Max
- Qwen3.8-Max
View all scores as a table
| Benchmark | Opus 4.8 | Fable 5 | GPT-5.6 Sol max | Qwen3.7-Max | Qwen3.8-Max |
|---|---|---|---|---|---|
| Coding Agent | |||||
| Terminal Bench 2.1 | 84.6 | 84.6 | 88.8 | 74.5 | 86.6 |
| SWE-bench Pro | 69.2 | 80.0 | 64.6 | 60.6 | 67.7 |
| DeepSWE 1.1 | 59.0 | 70.0 | 73.0 | 21.6 | 56.6 |
| NL2Repo-Bench | 69.4 | n/a | n/a | 47.2 | 55.9 |
| FrontierSWE | 70.0 | 88.8 | n/a | 40.7 | 73.5 |
| MLS-Bench-Lite | 42.8 | 49.9 | 46.2 | 31.7 | 41.0 |
| PaperBench | 80.3 | 88.8 | 90.5 | 64.8 | 93.0 |
| AndroidBench | 69.8 | 84.5 | 74.0 | 56.5 | 75.1 |
| QwenSWEBench | 84.0 | 86.3 | 73.5 | 63.4 | 80.7 |
| QwenQoderBench | 62.7 | 63.1 | 53.8 | 36.8 | 58.4 |
| QwenReactBench (Elo) | 1,694 | 1,770 | 1,564 | 1,538 | 1,724 |
| QwenSVGBench (Elo) | 1,648 | 1,690 | 1,758 | 1,499 | 1,713 |
| General Agent | |||||
| CoWorkBench | 72.3 | 75.9 | 71.5 | 64.6 | 74.8 |
| WorkSpaceBench | 66.8 | 68.7 | 65.6 | 61.4 | 67.7 |
| JobBench | 48.4 | 57.4 | 45.4 | 31.3 | 53.4 |
| SkillsBench | 65.1 | 70.9 | 73.5 | 61.2 | 70.2 |
| Agents' Last Exam (Score) | 45.1 | n/a | 53.6 | 31.1 | 52.4 |
| Automation-Bench (Pass@1) | 27.2 | 29.1 | 29.7 | 14.2 | 27.3 |
| Toolathlon Verified (Pass@1) | 76.2 | 77.9 | 74.9 | 49.7 | 72.5 |
| WideSearch (Item-F1) | 72.9 | 81.2 | n/a | 75.2 | 81.9 |
| HLE w/ tools | 57.9 | 64.5 | 58.0 | 53.5 | 56.2 |
| General Capabilities | |||||
| GPQA Diamond | 92.0 | 92.6 | 94.1 | 92.4 | 92.6 |
| HLE | 45.7 | 53.3 | 47.2 | 41.4 | 43.6 |
| IFBench | 62.2 | 63.5 | 72.7 | 79.1 | 82.8 |
| $OneMillion-Bench (Expert score) | 41.8 | 55.9 | 53.8 | 44.4 | 52.5 |
| HealthBench | 52.4 | n/a | 55.3 | 54.5 | 60.2 |
| PLawBench | 69.6 | 70.2 | 72.3 | 58.9 | 73.2 |
| PRBench-Legal | 52.7 | 57.6 | 57.6 | 48.5 | 57.6 |
| PRBench-Finance | 51.9 | 55.8 | 55.5 | 46.8 | 58.3 |
| Long Context | |||||
| MRCR v2 256K (8-needle) | 83.2 | n/a | 93.8 | 86.7 | 92.9 |
| LongBench v2 | 69.1 | n/a | 67.1 | 65.3 | 66.3 |
Source: the Qwen3.8-Max model card. Best score in each row is marked.
Third-party evaluations
Independent leaderboards started scoring Qwen3.8-Max within a day of the API launch. The picture is consistent across coding, design, and vision boards: a top-three finish behind the Claude frontier models, at a much lower list price.
API pricing
Qwen3.8-Max is served through QwenCloud at $2.00 per million input tokens and $6.00 per million output tokens. One flat rate covers the whole 1M-token context window, so there is no long-prompt surcharge. Implicit cache reads drop input to $0.25 per million tokens, an 8x discount. Explicit caching costs $2.50 per million tokens to create and $0.17 per million tokens to read.
| Model | Input · cache hit | Input · cache miss | Output |
|---|---|---|---|
| Qwen3.8-Max | $0.25 | $2.00 | $6.00 |
Per 1M tokens, as of August 6, 2026. Official pricing.
That output price undercuts the closed frontier by a wide margin. Claude Opus 5 costs $25 per million output tokens, so Qwen3.8-Max is roughly 76% cheaper. Claude Fable 5 costs $50, a 88% gap. GPT-5.6 Sol costs $30, an 80% gap. The Arena Pareto chart puts the model on the price-performance frontier at $5.00 per million blended tokens.
Sources: QwenCloud pricing · Claude pricing · OpenAI pricing
What people are saying
“We ran a test between the new Qwen3.8-Max, Opus 5 and GPT-5.6 Sol. 3 models. same prompt. one-shot with the /design command. Reviewed gameplay features, UX/UI and cost. 🔹 Qwen3.8-Max → 9/10 · $0.0248 🔹 GPT-5.6 Sol → 9/10 · $0.150 🔹 Opus 5 → 8.5/10 · $0.253 Qwen3.8-Max is approximately 4.2× cheaper than GPT-5.6, Sol, Opus 5 and has the same level of UI, UX, and gameplay.”
Command Code
@CommandCodeAI
X Aug 3, 2026
“Qwen3.8-Max is the best object detection VLM - satellite images - infrared images - documents - techical drawings - hand-sketched diagrams - crowded scenes - small objects”
SkalskiP
@skalskip92
X Aug 3, 2026
“We ran Qwen3.8-Max on our cybersecurity benchmark. Given enough attempts, it found more CVEs than most frontier models, tying Opus 5 for first place... Across the three runs, Qwen found 26 of 32 CVEs, reaching 81.25% pass@3 recall, outperforming GPT-5.6-Sol and matching Opus 5 at a lower price... Qwen is very inconsistent, but can be very strong if run multiple times.”
pilvar (Philippe Dourassov)
@pilvar222 · Aikido Security
X Aug 4, 2026
“It looks like Qwen3.8-Max has a major overthinking problem, it uses a LOT of tokens and is very slow, period, no other way to see it. My cost to run this benchmark was $126.25, to run the exact same eval suite with DeepSeek V4-Flash was only $13.60. This makes Qwen3.8-Max an insanely expensive model.”
Morgan
@morganlinton
X Aug 4, 2026
“We ran a test between the new Qwen3.8-Max, Opus 5 and GPT-5.6 Sol. 3 models. same prompt. one-shot with the /design command. Reviewed gameplay features, UX/UI and cost. 🔹 Qwen3.8-Max → 9/10 · $0.0248 🔹 GPT-5.6 Sol → 9/10 · $0.150 🔹 Opus 5 → 8.5/10 · $0.253 Qwen3.8-Max is approximately 4.2× cheaper than GPT-5.6, Sol, Opus 5 and has the same level of UI, UX, and gameplay.”
Command Code
@CommandCodeAI
X Aug 3, 2026
“Qwen3.8-Max is the best object detection VLM - satellite images - infrared images - documents - techical drawings - hand-sketched diagrams - crowded scenes - small objects”
SkalskiP
@skalskip92
X Aug 3, 2026
“We ran Qwen3.8-Max on our cybersecurity benchmark. Given enough attempts, it found more CVEs than most frontier models, tying Opus 5 for first place... Across the three runs, Qwen found 26 of 32 CVEs, reaching 81.25% pass@3 recall, outperforming GPT-5.6-Sol and matching Opus 5 at a lower price... Qwen is very inconsistent, but can be very strong if run multiple times.”
pilvar (Philippe Dourassov)
@pilvar222 · Aikido Security
X Aug 4, 2026
“It looks like Qwen3.8-Max has a major overthinking problem, it uses a LOT of tokens and is very slow, period, no other way to see it. My cost to run this benchmark was $126.25, to run the exact same eval suite with DeepSeek V4-Flash was only $13.60. This makes Qwen3.8-Max an insanely expensive model.”
Morgan
@morganlinton
X Aug 4, 2026
Frequently asked questions
Is Qwen3.8-Max open source?
Not yet. Alibaba announced on August 3, 2026 that Qwen3.8-Max is the first Qwen model at Max scale that will be open-sourced, and said the weights will land on Hugging Face and ModelScope the following week. As of August 6, 2026 the weights are not published.
How big is Qwen3.8-Max?
Qwen3.8-Max has 2.4 trillion total parameters and activates 95 billion per token. It is a Mixture-of-Experts model built on the architectural foundation of Qwen3.5. Alibaba has not published the expert count, the attention design, the vocabulary size, or the training precision.
What is the Qwen3.8-Max context window?
One million tokens. The QwenCloud API accepts up to 991K input tokens, or 983K with thinking enabled, and returns up to 131K output tokens. The reasoning budget can run to 262K tokens. Pricing is flat across the whole window, with no long-prompt surcharge.
How much does the Qwen3.8-Max API cost?
On QwenCloud it costs $2.00 per million input tokens and $6.00 per million output tokens. Implicit cache reads cost $0.25 per million tokens. Explicit cache creation costs $2.50 per million tokens and explicit cache reads cost $0.17 per million tokens.
How does Qwen3.8-Max compare to Claude Fable 5 and GPT-5.6 Sol?
It trades wins with both. Qwen3.8-Max leads the group on PaperBench at 93.0, IFBench at 82.8, WideSearch at 81.9, and PLawBench at 73.2. GPT-5.6 Sol leads Terminal Bench 2.1 at 88.8 against 86.6, and Claude Fable 5 leads SWE-bench Pro at 80.0 against 67.7 and FrontierSWE at 88.8 against 73.5. The list price gap is large: $6 per million output tokens against $30 for GPT-5.6 Sol and $50 for Claude Fable 5.