Compare models

Context, limits, capabilities, price and use cases side by side — up to 4 models at a time. The URL holds your selection, so you can share it.

Qwen3.8 MaxQwen3.7 Max Add a modelChat with these 2

AttributeHide identical rows	Qwen3.8 Maxqwen/qwen3.8-max	Qwen3.7 Maxqwen/qwen3.7-max20% off
Specs
Context window	1M	1M
Max output	131K	66K
Input modalities	TextImageVideo	Text
Output modalities	Text	Text
Capabilities
Streaming	Supported	Supported
Tool calling	Supported	Supported
Structured output (JSON)	Supported	Supported
Image input	Supported	Not supported
Thinking mode	Supported	Supported
Price / 1M tokens
Input	$ 1.88	$ 1.88$ 1.50
Output	$ 5.63	$ 5.63$ 4.50
Cache read	$ 0.23	$ 0.38$ 0.30
Cache write	$ 2.35	$ 2.35$ 1.88
What it is good at
Best for	Maximum reasoning Multimodal coding agents	Complex reasoning Long-running coding agents
Published benchmarks
Terminal-Bench 2.0pass_rate	—	69.7%Terminus(setup not fully disclosed)
Terminal-Bench 2.1pass_rate	86.6%Claude Code · 10 runs(setup not fully disclosed)	—
SWE-Bench Proresolved_ratesetups differ by: harness, temperature, top_p	67.7%Claude Code(setup not fully disclosed)	60.6%(setup not fully disclosed)
SWE-Bench Verifiedresolved_rate	—	80.4%(setup not fully disclosed)
OSWorld Verifiedsuccess_rate	86.1%(setup not fully disclosed)	—
IFBenchscore	—	79.1%(setup not fully disclosed)
MRCR v2 128Kscore	—	90.4%(setup not fully disclosed)
MRCR v2 256K (8-needle)score	92.9%(setup not fully disclosed)	—
MMMU Proaccuracy	82.3%(setup not fully disclosed)	—
Humanity's Last Examaccuracy	43.6%(setup not fully disclosed)	41.4%(setup not fully disclosed)
Humanity's Last Exam (with tools)accuracy	—	53.5%(setup not fully disclosed)
PaperBench (BasicAgent)score	93%3 runs(setup not fully disclosed)	—
GPQA Diamondaccuracy	92.6%(setup not fully disclosed)	92.4%(setup not fully disclosed)
MCPMarkscore	—	60.8%(setup not fully disclosed)
Toolathlon Verifiedpass_rate	72.5%(setup not fully disclosed)	—
VideoMMMUaccuracy	88.7%(setup not fully disclosed)	—