Hy3-preview vs Qwen3.7 Max
Wins 0 of 7 areas
—
Wins 4 of 7 areas
Coding · Reasoning · Facts · Following instructions
Qwen3.7 Max wins more areas, narrowly.Hy3-preview is cheaper.
Scores updated · 29 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. 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 software05Qwen3.7 Max5 of 7 tests · 2 ties
- ReasoningHard problems that need careful thinking03Qwen3.7 Max3 of 3 tests
- FactsGetting facts right instead of making them up01Qwen3.7 Max1 of 2 tests · 1 tie
- Following instructionsDoing exactly what it is asked01Qwen3.7 Max1 of 1 test
- AgentsCarrying out multi-step tasks on its own22Even2 each
- MathCompetition and research-level math00Even0 each · 1 tie
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
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.
Hy3-preview costs 93% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Qwen3.7 Max pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+48.5points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+32.5points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+19points ahead
Where Hy3-preview pulls ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+11.1points ahead
- Multi-step tasks using many real software toolsMCP Atlas+2.7points ahead
Every test, side by side
All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3.7 Max
- Terminal-Bench HardQwen3.7 Max by 1931.850.8+19
- Terminal-Bench 2.1Qwen3.7 Max by 10.164.474.5+10.1
- SWE-bench ProQwen3.7 Max by 2.757.960.6+2.7
- SWE-bench MultilingualQwen3.7 Max by 2.575.878.3+2.5
- SWE-bench VerifiedQwen3.7 Max by 2.47880.4+2.4
- SciCodetie48.649.5tie
- LMArena · WebDevtie15081515tie
AgentsEven
- τ-Bench V3 · BankingHy3-preview by 11.122.911.8+11.1
- GDPValQwen3.7 Max by 3.328.331.6+3.3
- MCP AtlasHy3-preview by 2.779.176.4+2.7
- Terminal-Bench 4.0Qwen3.7 Max by 10.51.5+1
ReasoningQwen3.7 Max
- CritPtQwen3.7 Max by 8.54.913.4+8.5
- Humanity's Last ExamQwen3.7 Max by 733.540.5+7
- GPQA DiamondQwen3.7 Max by 2.689.792.3+2.6
FactsQwen3.7 Max
- AA-Omniscience · Non-hallucinationQwen3.7 Max by 48.525.974.4+48.5
- AA-Omniscience · Accuracytie3231.1tie
Following instructionsQwen3.7 Max
- IFBenchQwen3.7 Max by 32.54880.5+32.5
Other results10 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-OmniscienceQwen3.7 Max by 32-18.513.5+32
- τ²-Bench Telecom (AA run)Qwen3.7 Max by 27.267.594.7+27.2
- DeepSWEHy3-preview by 102818+10
- Artificial Analysis Coding IndexQwen3.7 Max by 7.258.866+7.2
- MathArena Apex (Pass@1)Qwen3.7 Max by 5.838.744.5+5.8
- AA IntelligenceQwen3.7 Max by 4.225.329.5+4.2
- Claw Eval (pass@3)Hy3-preview by 3.368.565.2+3.3
- AA Agentic IndexHy3-preview by 1.725.623.9+1.7
- NL2RepoQwen3.7 Max by 1.645.647.2+1.6
- WideSearchHy3-preview by 1.276.475.2+1.2
Questions people ask
Which is better, Hy3-preview or Qwen3.7 Max?
Qwen3.7 Max wins four of the seven areas where both have results: coding, reasoning, facts and following instructions. Hy3-preview wins none, but costs 93% less. They are level on agents, math and long documents.
Which is better for coding?
Qwen3.7 Max. It wins 5 of the 7 coding tests both models report; Hy3-preview wins none, and 2 are ties.
Which is cheaper?
Hy3-preview costs $0.14 per million input tokens and $0.56 per million output tokens; Qwen3.7 Max costs $2.50 and $7.50. That makes Hy3-preview about 93% cheaper for the same work.
How do you compare the two?
We use the 29 benchmark tests both models have published scores on. The verdict counts the 19 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 10 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.