MODEL PROFILE / EDITION 2026.09
Qwen3.7 Max
High-throughput algorithmic and component-level code generation where speed matters
Balanced model evidence score; rank 21. No rank delta is shown because there is no prior numeric snapshot.
Best for
High-throughput algorithmic and component-level code generation where speed matters
Price & access
$2.50 in / $7.50 out per 1M (Artificial Analysis); supports explicit cache
Evidence summary
Arena, Design Arena and three third-party benchmark boards cover it, but pricing and context come from Artificial Analysis rather than a fetched Alibaba price row, and its LiveCodeBench board is very small.
Edition status
No prior snapshot. Tracked in the September 2026 baseline.
Strengths
- LiveCodeBench rank 1 (91.6%) - the strongest competitive-coding evidence in this set
- 202.3 tok/s with 2.29 s TTFT: one of the fastest frontier-class models measured
- SWE-bench Verified 80.4% at $2.50/$7.50
- Well-sampled Arena row (7,281 votes)
Tradeoffs
- Arena frontend 1520 and Design Arena 1289 are mid-pack; Qwen3.8 Max is 180 Elo higher
- LiveCodeBench's board only has 7 models, so 'rank 1' is thin
- AA Intelligence Index 37 is low for the price
- Vision only arrived in a later snapshot, so multimodal behaviour depends on the snapshot pinned
Score profile
visual 65 · code 76 · agentic 72 · debug 70 · context 86 · speed 92 · value 76
Linked evidence
Evidence source 1 ↗Evidence source 2 ↗Evidence source 3 ↗Evidence source 4 ↗Evidence source 5 ↗Evidence source 6 ↗Use the same published data, filters, and side-by-side metrics.