GPT-5 vs Qwen3.5 397B A17B
Wins 2 of 8 areas
Agents · Facts
Wins 2 of 8 areas
Coding · Images and charts
The two are evenly matched.GPT-5 is better at facts and agents; Qwen3.5 397B A17B at coding and images and charts.
Scores updated · 39 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software26Qwen3.5 397B A17B6 of 8 tests
- Images and chartsUnderstanding pictures, charts and video02Qwen3.5 397B A17B2 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up30GPT-53 of 3 tests
- AgentsCarrying out multi-step tasks on its own21GPT-52 of 3 tests
- ReasoningHard problems that need careful thinking11Even1 each · 1 tie
- MathCompetition and research-level math11Even1 each
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
- Following instructionsDoing exactly what it is asked11Even1 each
Long documents rests on a single test.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Qwen3.5 397B A17B costs 63% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Qwen3.5 397B A17B pulls ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+16.1points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+14.1points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+14points ahead
Where GPT-5 pulls ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+24.2points ahead
- Short factual questions, answered correctlySimpleQA Verified+24.1points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+9.5points ahead
Every test, side by side
All 39 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3.5 397B A17B
- Terminal-Bench 2.1Qwen3.5 397B A17B by 16.135.251.3+16.1
- SWE-bench MultilingualQwen3.5 397B A17B by 1455.369.3+14
- SWE-bench ProQwen3.5 397B A17B by 9.141.850.9+9.1
- Terminal-Bench HardQwen3.5 397B A17B by 8.332.640.9+8.3
- LiveCodeBench v6GPT-5 by 3.48783.6+3.4
- SciCodeQwen3.5 397B A17B by 1.942.944.8+1.9
- LMArena · WebDevGPT-5 by 18 rating points14181400+18 rating
- SWE-bench VerifiedQwen3.5 397B A17B by 1.574.976.4+1.5
AgentsGPT-5
- BrowseCompQwen3.5 397B A17B by 14.154.969+14.1
- τ-Bench V3 · BankingGPT-5 by 8.722.113.4+8.7
- GDPValGPT-5 by 6.22114.8+6.2
ReasoningEven
- CritPtGPT-5 by 45.71.7+4
- GPQA DiamondQwen3.5 397B A17B by 3.985.489.3+3.9
- Humanity's Last Examtie28.529tie
FactsGPT-5
- SimpleQA VerifiedGPT-5 by 24.150.126+24.1
- AA-Omniscience · AccuracyGPT-5 by 9.540.330.8+9.5
- AA-Omniscience · Non-hallucinationGPT-5 by 6.717.811.1+6.7
Images and chartsQwen3.5 397B A17B
- LMArena · VisionQwen3.5 397B A17B by 56 rating points12081264+56 rating
- MMMU-ProQwen3.5 397B A17B by 3.174.277.3+3.1
- MMMUtie84.285tie
MathEven
- FrontierMath Tiers 1-3 (v2)GPT-5 by 24.255.431.2+24.2
- IMOAnswerBenchQwen3.5 397B A17B by 4.97680.9+4.9
Following instructionsEven
- IFBenchQwen3.5 397B A17B by 5.773.178.8+5.7
- Multi-ChallengeGPT-5 by 269.667.6+2
Other results14 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- AA-OmniscienceGPT-5 by 22-8.7-30.8+22
- Terminal-Bench 2.0Qwen3.5 397B A17B by 17.335.252.5+17.3
- AA Agentic IndexGPT-5 by 15.926.510.6+15.9
- HLE (with tools)Qwen3.5 397B A17B by 13.135.248.3+13.1
- τ²-Bench Telecom (AA run)Qwen3.5 397B A17B by 10.884.895.6+10.8
- Artificial Analysis Coding IndexQwen3.5 397B A17B by 10.437.848.2+10.4
- BrowseComp-ZHQwen3.5 397B A17B by 7.36370.3+7.3
- HMMT Feb. 2025Qwen3.5 397B A17B by 6.588.394.8+6.5
- AA IntelligenceGPT-5 by 4.62318.4+4.6
- HMMT Nov. 2025Qwen3.5 397B A17B by 3.589.292.7+3.5
- GPQA (unspecified)Qwen3.5 397B A17B by 1.485.787.1+1.4
- MMLU-Protie87.187.8tie
- HMMT 2025tie93.392.7tie
- MMLU-Reduxtie95.394.9tie
Questions people ask
Which is better, GPT-5 or Qwen3.5 397B A17B?
GPT-5 and Qwen3.5 397B A17B each win two of the eight areas we test. GPT-5 wins agents and facts; Qwen3.5 397B A17B wins coding and images and charts. Qwen3.5 397B A17B costs 63% less. They are level on reasoning, math, long documents and following instructions.
Which is better for coding?
Qwen3.5 397B A17B. It wins 6 of the 8 coding tests both models report; GPT-5 wins 2.
Which is cheaper?
GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens; Qwen3.5 397B A17B costs $0.60 and $3.60. That makes Qwen3.5 397B A17B about 63% cheaper for the same work.
How do you compare the two?
We use the 39 benchmark tests both models have published scores on. The verdict counts the 25 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 14 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.