GPT-5.6 Luna vs Qwen3.8 27B
Wins 5 of 8 areas
Coding · Agents · Reasoning · Math · Long documents
Wins 2 of 8 areas
Images and charts · Following instructions
GPT-5.6 Luna is the stronger all-rounder.Qwen3.8 27B is better at images and charts.
Scores updated · 41 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 software52GPT-5.6 Luna5 of 7 tests
- ReasoningHard problems that need careful thinking41GPT-5.6 Luna4 of 6 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own32GPT-5.6 Luna3 of 5 tests
- MathCompetition and research-level math10GPT-5.6 Luna1 of 1 test
- Long documentsFinding answers in very long texts10GPT-5.6 Luna1 of 1 test
- Images and chartsUnderstanding pictures, charts and video12Qwen3.8 27B2 of 3 tests
- Following instructionsDoing exactly what it is asked01Qwen3.8 27B1 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
Math, long documents and following instructions 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.
GPT-5.6 Luna costs 60% 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-5.6 Luna pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+27.1points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+21.7points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+17.2points ahead
Where Qwen3.8 27B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+62.3points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+16.9points ahead
- Common-sense trick questionsSimpleBench+13.4points ahead
Every test, side by side
All 41 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.6 Luna
- SWE-bench VerifiedGPT-5.6 Luna by 15.29377.8+15.2
- LiveBench · Agentic CodingQwen3.8 27B by 1348.461.4+13
- LiveBench · CodingGPT-5.6 Luna by 7.282.975.7+7.2
- LMArena · WebDevQwen3.8 27B by 71 rating points15191590+71 rating
- SciCodeGPT-5.6 Luna by 753.646.6+7
- Terminal-Bench 2.1GPT-5.6 Luna by 1.180.979.8+1.1
- SWE-bench ProGPT-5.6 Luna by 162.761.7+1
AgentsGPT-5.6 Luna
- BrowseCompGPT-5.6 Luna by 21.783.361.6+21.7
- τ-Bench V3 · BankingQwen3.8 27B by 16.931.148+16.9
- OSWorld-VerifiedQwen3.8 27B by 11.772.684.3+11.7
- Terminal-Bench 4.0GPT-5.6 Luna by 611.65.6+6
- GDPValGPT-5.6 Luna by 24846+2
ReasoningGPT-5.6 Luna
- ARC-AGI-2GPT-5.6 Luna by 17.259.642.4+17.2
- CritPtGPT-5.6 Luna by 15.220.65.4+15.2
- SimpleBenchQwen3.8 27B by 13.446.860.2+13.4
- Humanity's Last ExamGPT-5.6 Luna by 5.639.533.9+5.6
- LiveBench · ReasoningGPT-5.6 Luna by 5.685.680+5.6
- GPQA Diamondtie91.190.5tie
FactsEven
- AA-Omniscience · Non-hallucinationQwen3.8 27B by 62.37.469.7+62.3
- AA-Omniscience · AccuracyGPT-5.6 Luna by 27.142.715.6+27.1
Images and chartsQwen3.8 27B
- CharXiv (RQ)Qwen3.8 27B by 7.582.790.2+7.5
- MMMU-ProGPT-5.6 Luna by 2.378.676.3+2.3
- LMArena · VisionQwen3.8 27B by 15 rating points12561271+15 rating
Following instructionsQwen3.8 27B
- LiveBench · Instruction FollowingQwen3.8 27B by 12.660.172.7+12.6
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.
- DeepSWE 1.1GPT-5.6 Luna by 24.86742.2+24.8
- LVBenchQwen3.8 27B by 13.558.972.4+13.5
- RealWorldQAQwen3.8 27B by 13.572.485.9+13.5
- Terminal-Bench 4.0GPT-5.6 Luna by 11.717.35.6+11.7
- ERQAQwen3.8 27B by 956.565.5+9
- Agents' Last ExamGPT-5.6 Luna by 7.450.342.9+7.4
- GDP (Surge AI)GPT-5.6 Luna by 6.322.716.4+6.3
- AA Agentic IndexQwen3.8 27B by 3.842.746.5+3.8
- AA IntelligenceGPT-5.6 Luna by 3.637.333.7+3.6
- Artificial Analysis Coding IndexGPT-5.6 Luna by 3.371.468.1+3.3
- ARC-AGI-1GPT-5.6 Luna by 3.290.787.5+3.2
- livebench_languageQwen3.8 27B by 1.872.674.3+1.8
- livebench_data_analysisGPT-5.6 Luna by 1.47876.6+1.4
- Toolathlon Verifiedtie67.967.1tie
- AA-Omnisciencetie-10.3-10tie
Questions people ask
Which is better, GPT-5.6 Luna or Qwen3.8 27B?
GPT-5.6 Luna wins five of the eight areas we test: coding, agents, reasoning, math and long documents. Qwen3.8 27B wins images and charts and following instructions. They are level on facts.
Which is better for coding?
GPT-5.6 Luna. It wins 5 of the 7 coding tests both models report; Qwen3.8 27B wins 2.
Which is cheaper?
GPT-5.6 Luna costs $0.20 per million input tokens and $1.20 per million output tokens; Qwen3.8 27B costs $0.50 and $3.00. That makes GPT-5.6 Luna about 60% cheaper for the same work.
How do you compare the two?
We use the 41 benchmark tests both models have published scores on. The verdict counts the 26 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.