Skip to content
Alpha
  • Models
  • Rankings
  • Ori

Footer

OpenRouterOpenRouter
© 2026 OpenRouter, Inc

Product

  • Chat
  • Rankings
  • Benchmarks
  • Apps
  • Discover
  • Models
  • Ori
  • Collections
  • Providers
  • Tools
  • Pricing
  • Business
  • Enterprise
  • Labs

Company

  • About
  • Blog
  • Customers
  • Careers
    Hiring
  • Privacy
  • Terms of Service
  • Trust Center
  • Support
  • Works With OR
  • Data
  • Brand

Developer

  • Documentation
  • API Reference
  • Developer Platform
  • Status
  • AI Site Map

Connect

  • Discord
  • GitHub
  • LinkedIn
  • X
  • YouTube
Favicon for qwen

Qwen: Qwen3 235B A22B Thinking 2507

qwen/qwen3-235b-a22b-thinking-2507

Model weights
Compare

Qwen3-235B-A22B-Thinking-2507 is a high-performance, open-weight Mixture-of-Experts (MoE) language model optimized for complex reasoning tasks. It activates 22B of its 235B parameters per forward pass and natively supports up to 262,144 tokens of context. This "thinking-only" variant enhances structured logical reasoning, mathematics, science, and long-form generation, showing strong benchmark performance across AIME, SuperGPQA, LiveCodeBench, and MMLU-Redux. It enforces a special reasoning mode (</think>) and is designed for high-token outputs (up to 81,920 tokens) in challenging domains.

The model is instruction-tuned and excels at step-by-step reasoning, tool use, agentic workflows, and multilingual tasks. This release represents the most capable open-source variant in the Qwen3-235B series, surpassing many closed models in structured reasoning use cases.

Modalities
In / Out Price
$0.45 / $3.50per 1M
Context
128K
Released
Jul 25, 2025
Knowledge Cutoff
Jun 2025
Compare
ProvidersPricingPerformanceUptimeBenchmarksAppsActivityFAQExplore

Providers

This model is hosted by one provider. OpenRouter forwards every request to it directly — no routing decisions to make.

Pricing

The average price customers actually pay for this model, next to the prices providers post. Caching and discounts mean the price actually paid is often well below the listed one.

Performance

Throughput is how fast the model writes (tokens per second — higher is better). Latency is total round-trip time (lower is better). TTFT is time-to-first-token — how long before you see anything appear (lower is better).

Uptime

Uptime is the percentage of the past 3 days that at least one provider was responding to requests. Availability is the percentage of time that inference was successfully served. OpenRouter continuously monitors and uses the next-best provider when one returns an error.

Benchmarks

Scores on standardized evaluations and this model’s own outputs on shared benchmark prompts. Higher percentages are better — and each rank shows where this model lands among the models evaluated on that source.

Benchmark score summary for Qwen: Qwen3 235B A22B Thinking 2507 (Artificial Analysis and Design Arena)
SourceBenchmarkScore
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) Intelligence Index12.7
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) Coding Index22.1
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) Math Index91.0
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) Agentic Index1.3
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) GPQA Diamond79.0%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) HLE15.9%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) IFBench51.2%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) τ²-Bench Telecom53.2%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) AA-LCR72.0%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) τ-Bench Banking7.8%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) GDPval-AA0.0%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) CritPt0.0%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) LiveCodeBench78.8%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) SciCode41.4%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) Terminal-Bench Hard13.6%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) Terminal-Bench 4.00.0%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) Terminal-Bench 2.112.0%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) MMLU-Pro84.3%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) AA-Omniscience Accuracy22.8%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) AA-Omniscience Non-Hallucination Rate10.2%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) AIME 202591.0%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) AIME94.0%
Artificial AnalysisQwen3 235B A22B 2507 (Reasoning) MATH-50098.4%
Design ArenaQwen3-235B-A22B-Thinking-2507 Models Arena 3D Elo995
Design ArenaQwen3-235B-A22B-Thinking-2507 Models Arena Code Categories Elo1031
Design ArenaQwen3-235B-A22B-Thinking-2507 Models Arena Data Visualization Elo950
Design ArenaQwen3-235B-A22B-Thinking-2507 Models Arena Game Development Elo957
Design ArenaQwen3-235B-A22B-Thinking-2507 Models Arena UI Component Elo940
Design ArenaQwen3-235B-A22B-Thinking-2507 Models Arena Website Elo1051

Apps

Public apps that send the most traffic to this model. Good signal for what real production workloads look like — and a hint at which use cases this model is best suited for.

Activity

Token volume and request traffic to this model over time.

API

Drop-in code to call this model. OpenRouter's API is OpenAI-compatible — most SDKs work by just swapping the base URL. The only thing that changes between models is the model slug below.

Not currently available through EU in-region routing. In-region routing docs

Explore more models

AI Model RankingsRanking
Latency / throughput
$0.45$3.500.68s94 tps
91.90%
Throughput
94tok/s

P50, best across providers

Latency
0.68s

P50, best provider

Uptime (3d)The model was reachable. Request routed to a provider.
100.00%
Availability (3d)The model returned inference from any provider. Errors and empty responses count against it.
94.68%

Availability over the last 3 days

Last 72 hours
Availability 94.68%
3 Days Ago2 Days AgoYesterdayNow

Availability over the last 24 hours

OpenRouter Availability
93.72%

When an error occurs in an upstream provider, we can recover by routing to another healthy provider, if your request filters allow it. You can access per-provider uptime data programmatically through the Endpoints API. Learn more about our load balancing and provider routing options.

1.
Favicon for https://nousresearch.com
Hermes Agent
Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across sessions, and builds reusable skills from experience. It comes with 40+ built-in tools, including web search, browser automation, and vision, plus scheduled automations and subagents.
240Mtokens
2.
Favicon for https://baseten.com/
Baseten takehome evaluation
new
57.9Mtokens
3.
Favicon for https://omp.sh/
omp
new
31.5Mtokens
4.
Favicon for https://openclaw.ai/
OpenClaw
OpenClaw is an open-source AI agent that connects to your messaging apps and takes real actions on your behalf, from running commands and browsing the web to managing files and sending emails.
30Mtokens
5.
Favicon for https://ufal.mff.cuni.cz/
can-llms-count-e04
new
28.5Mtokens

Frequently asked questions

Qwen3-235B-A22B-Thinking-2507 is a high-performance, open-weight Mixture-of-Experts (MoE) language model optimized for complex reasoning tasks. It activates 22B of its 235B parameters per forward pass and natively supports up to 262,144 tokens of context.

Qwen3 235B A22B Thinking 2507 costs $0.45/M input tokens and $3.50/M output tokens.

Qwen3 235B A22B Thinking 2507 has a 128,000 token context window. It supports up to 16,384 completion tokens.

Yes. Qwen3 235B A22B Thinking 2507 accepts tools and tool_choice for function calling. It does not support response_format, so JSON output is not enforced.

Qwen3.8 Max Prime, Qwen3.8 Omni Flash, Qwen3.8 Max (0902) and 38 more are other text models from Qwen.

Qwen3 235B A22B Thinking 2507 was released on July 25, 2025. Its knowledge cutoff is June 30, 2025.