GPT-5.5 vs Qwen3.8 27B
Wins 4 of 8 areas
Reasoning · Images and charts · Math · Long documents
Wins 2 of 8 areas
Agents · Following instructions
GPT-5.5 wins more areas, narrowly.Qwen3.8 27B is cheaper and better at agents.
Scores updated · 40 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.
- ReasoningHard problems that need careful thinking60GPT-5.56 of 6 tests
- Images and chartsUnderstanding pictures, charts and video21GPT-5.52 of 3 tests
- MathCompetition and research-level math10GPT-5.51 of 1 test
- Long documentsFinding answers in very long texts10GPT-5.51 of 1 test
- AgentsCarrying out multi-step tasks on its own13Qwen3.8 27B3 of 4 tests
- Following instructionsDoing exactly what it is asked02Qwen3.8 27B2 of 2 tests
- CodingWriting and fixing software33Even3 each
- FactsGetting facts right instead of making them up11Even1 each
Math and long documents rest on a single test each.
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.
Qwen3.8 27B costs 90% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-5.5 pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+42.6points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+42.4points ahead
- Research-level physics problemsCritPt+21.7points ahead
Where Qwen3.8 27B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+58.7points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+9points ahead
- Fixes real GitHub issues, working as an agentLiveBench · Agentic Coding+7.4points ahead
Every test, side by side
All 40 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- SciCodeGPT-5.5 by 9.255.846.6+9.2
- LMArena · WebDevQwen3.8 27B by 79 rating points15131592+79 rating
- LiveBench · Agentic CodingQwen3.8 27B by 7.45461.4+7.4
- LiveBench · CodingGPT-5.5 by 6.582.275.7+6.5
- Terminal-Bench 2.1GPT-5.5 by 4.584.379.8+4.5
- SWE-bench ProQwen3.8 27B by 3.158.661.7+3.1
AgentsQwen3.8 27B
- Terminal-Bench 4.0GPT-5.5 by 914.65.6+9
- τ-Bench V3 · BankingQwen3.8 27B by 93948+9
- OSWorld-VerifiedQwen3.8 27B by 5.678.784.3+5.6
- GDPValQwen3.8 27B by 3.542.746.2+3.5
ReasoningGPT-5.5
- ARC-AGI-2GPT-5.5 by 42.68542.4+42.6
- CritPtGPT-5.5 by 21.727.15.4+21.7
- Humanity's Last ExamGPT-5.5 by 11.945.833.9+11.9
- LiveBench · ReasoningGPT-5.5 by 9.789.780+9.7
- SimpleBenchGPT-5.5 by 8.86960.2+8.8
- GPQA DiamondGPT-5.5 by 393.590.5+3
FactsEven
- AA-Omniscience · Non-hallucinationQwen3.8 27B by 58.71169.7+58.7
- AA-Omniscience · AccuracyGPT-5.5 by 42.45815.6+42.4
Images and chartsGPT-5.5
- MMMU-ProGPT-5.5 by 3.679.976.3+3.6
- OCRBenchv2Qwen3.8 27B by 2.461.163.5+2.4
- LMArena · VisionGPT-5.5 by 22 rating points12921270+22 rating
Following instructionsQwen3.8 27B
- IFBenchQwen3.8 27B by 3.675.979.5+3.6
- LiveBench · Instruction FollowingQwen3.8 27B by 270.772.7+2
Other results15 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-5.5 by 30.520.5-10+30.5
- BabyVisionQwen3.8 27B by 29.755.985.6+29.7
- DeepSWE 1.1GPT-5.5 by 24.86742.2+24.8
- Agents' Last ExamQwen3.8 27B by 16.326.642.9+16.3
- livebench_languageGPT-5.5 by 1387.474.3+13
- AA Agentic IndexQwen3.8 27B by 9.237.346.5+9.2
- NL2RepoGPT-5.5 by 8.450.742.3+8.4
- ARC-AGI-1GPT-5.5 by 7.59587.5+7.5
- Artificial Analysis Coding IndexGPT-5.5 by 6.874.968.1+6.8
- Toolathlon VerifiedGPT-5.5 by 6.473.567.1+6.4
- livebench_data_analysisGPT-5.5 by 581.676.6+5
- JobBenchGPT-5.5 by 4.938.333.4+4.9
- AA IntelligenceGPT-5.5 by 4.738.433.7+4.7
- RealWorldQAQwen3.8 27B by 3.782.285.9+3.7
- ERQAQwen3.8 27B by 164.565.5+1
Questions people ask
Which is better, GPT-5.5 or Qwen3.8 27B?
GPT-5.5 wins four of the eight areas we test: reasoning, images and charts, math and long documents. Qwen3.8 27B wins agents and following instructions, and costs 90% less. They are level on coding and facts.
Which is better for coding?
Neither. They win 3 coding tests each of the 6 both models report.
Which is cheaper?
GPT-5.5 costs $5.00 per million input tokens and $30.00 per million output tokens; Qwen3.8 27B costs $0.50 and $3.00. That makes Qwen3.8 27B about 90% cheaper for the same work.
How do you compare the two?
We use the 40 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 15 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.