GPT-5.6 Luna vs Qwen3.8 Max Preview
Wins 2 of 8 areas
Math · Long documents
Wins 6 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Following instructions
Qwen3.8 Max Preview is the stronger all-rounder.GPT-5.6 Luna is cheaper and better at math.
Scores updated · 44 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.
- AgentsCarrying out multi-step tasks on its own05Qwen3.8 Max Preview5 of 6 tests · 1 tie
- CodingWriting and fixing software24Qwen3.8 Max Preview4 of 6 tests
- ReasoningHard problems that need careful thinking13Qwen3.8 Max Preview3 of 4 tests
- Images and chartsUnderstanding pictures, charts and video02Qwen3.8 Max Preview2 of 2 tests
- FactsGetting facts right instead of making them up12Qwen3.8 Max Preview2 of 3 tests
- Following instructionsDoing exactly what it is asked01Qwen3.8 Max Preview1 of 1 test
- MathCompetition and research-level math21GPT-5.6 Luna2 of 3 tests
- Long documentsFinding answers in very long texts10GPT-5.6 Luna1 of 1 test
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 83% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Qwen3.8 Max Preview pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+63.8points ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+27.3points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+16.7points ahead
Where GPT-5.6 Luna pulls ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+12.2points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+11points ahead
- Programming problems, refreshed regularlyLiveBench · Coding+10points ahead
Every test, side by side
All 44 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3.8 Max Preview
- LiveBench · Agentic CodingQwen3.8 Max Preview by 16.348.464.7+16.3
- LMArena · WebDevQwen3.8 Max Preview by 151 rating points15191670+151 rating
- LiveBench · CodingGPT-5.6 Luna by 1082.972.9+10
- Terminal-Bench 2.1Qwen3.8 Max Preview by 7.980.988.8+7.9
- SWE-bench ProQwen3.8 Max Preview by 562.767.7+5
- SciCodeGPT-5.6 Luna by 1.553.652.1+1.5
AgentsQwen3.8 Max Preview
- Terminal-Bench 4.0Qwen3.8 Max Preview by 27.311.638.9+27.3
- τ-Bench V3 · BankingQwen3.8 Max Preview by 16.731.147.8+16.7
- OSWorld-VerifiedQwen3.8 Max Preview by 13.572.686.1+13.5
- GDPValQwen3.8 Max Preview by 10.34858.3+10.3
- AA ApexAgentsQwen3.8 Max Preview by 6.635.842.4+6.6
- AA IT-Bench SREtie40.340.3tie
ReasoningQwen3.8 Max Preview
- Humanity's Last ExamQwen3.8 Max Preview by 3.639.543.1+3.6
- CritPtGPT-5.6 Luna by 2.920.617.7+2.9
- LiveBench · ReasoningQwen3.8 Max Preview by 2.685.688.2+2.6
- GPQA DiamondQwen3.8 Max Preview by 1.791.192.8+1.7
FactsQwen3.8 Max Preview
- AA-Omniscience · Non-hallucinationQwen3.8 Max Preview by 63.87.471.2+63.8
- AA-Omniscience · AccuracyGPT-5.6 Luna by 1142.731.7+11
- SimpleQA VerifiedQwen3.8 Max Preview by 6.34147.3+6.3
Images and chartsQwen3.8 Max Preview
- LMArena · VisionQwen3.8 Max Preview by 56 rating points12561312+56 rating
- MMMU-ProQwen3.8 Max Preview by 4.278.682.8+4.2
MathGPT-5.6 Luna
- FrontierMath Tier 4GPT-5.6 Luna by 12.258.546.3+12.2
- FrontierMath Tiers 1-3 (v2)GPT-5.6 Luna by 7.482.174.7+7.4
- LiveBench · MathematicsQwen3.8 Max Preview by 4.187.291.3+4.1
Following instructionsQwen3.8 Max Preview
- LiveBench · Instruction FollowingQwen3.8 Max Preview by 1460.174.1+14
Other results18 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.
- LVBenchQwen3.8 Max Preview by 22.958.981.8+22.9
- AA-OmniscienceQwen3.8 Max Preview by 22.3-10.312+22.3
- ERQAQwen3.8 Max Preview by 21.356.577.8+21.3
- GDPval-AA v2Qwen3.8 Max Preview by 209 rating points15301739+209 rating
- ToolathlonQwen3.8 Max Preview by 19.153.472.5+19.1
- RealWorldQAQwen3.8 Max Preview by 15.672.488+15.6
- DeepSWEGPT-5.6 Luna by 10.667.256.6+10.6
- DeepSWE 1.1GPT-5.6 Luna by 10.46756.6+10.4
- AA IntelligenceQwen3.8 Max Preview by 8.137.345.4+8.1
- HLE (with tools)Qwen3.8 Max Preview by 7.348.956.2+7.3
- livebench_languageQwen3.8 Max Preview by 7.172.679.7+7.1
- AA Agentic IndexQwen3.8 Max Preview by 6.942.749.6+6.9
- Toolathlon VerifiedQwen3.8 Max Preview by 4.667.972.5+4.6
- HealthBenchQwen3.8 Max Preview by 4.455.860.2+4.4
- Agents' Last ExamQwen3.8 Max Preview by 2.150.352.4+2.1
- Artificial Analysis Coding Indextie71.471.8tie
- livebench_data_analysistie7878.4tie
- GDP (Surge AI)tie22.722.8tie
Questions people ask
Which is better, GPT-5.6 Luna or Qwen3.8 Max Preview?
Qwen3.8 Max Preview wins six of the eight areas we test: coding, agents, reasoning, facts, images and charts and following instructions. GPT-5.6 Luna wins math and long documents, and costs 83% less.
Which is better for coding?
Qwen3.8 Max Preview. It wins 4 of the 6 coding tests both models report; GPT-5.6 Luna 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 Max Preview costs $2.00 and $6.00. That makes GPT-5.6 Luna about 83% cheaper for the same work.
How do you compare the two?
We use the 44 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 18 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.