GLM-5.2 vs Qwen3.6 35B A3B
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
Wins 0 of 7 areas
—
GLM-5.2 is the stronger all-rounder.Qwen3.6 35B A3B is cheaper.
Scores updated · 31 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 software40GLM-5.24 of 4 tests
- AgentsCarrying out multi-step tasks on its own40GLM-5.24 of 4 tests
- MathCompetition and research-level math40GLM-5.24 of 4 tests
- ReasoningHard problems that need careful thinking30GLM-5.23 of 3 tests
- FactsGetting facts right instead of making them up20GLM-5.22 of 2 tests
- Long documentsFinding answers in very long texts10GLM-5.21 of 1 test
- Following instructionsDoing exactly what it is asked10GLM-5.21 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.
Qwen3.6 35B A3B costs 55% 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 GLM-5.2 pulls ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+38.8points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+33points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+25.3points ahead
Where Qwen3.6 35B A3B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 31 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGLM-5.2
- Terminal-Bench 2.1GLM-5.2 by 3377.944.9+33
- Terminal-Bench HardGLM-5.2 by 1650.834.8+16
- SciCodeGLM-5.2 by 14.651.236.6+14.6
- SWE-bench ProGLM-5.2 by 12.662.149.5+12.6
AgentsGLM-5.2
- τ-Bench V3 · BankingGLM-5.2 by 25.334.69.3+25.3
- GDPValGLM-5.2 by 23.943.719.8+23.9
- MCP AtlasGLM-5.2 by 1476.862.8+14
- Terminal-Bench 4.0GLM-5.2 by 110+1
ReasoningGLM-5.2
- CritPtGLM-5.2 by 20.620.90.3+20.6
- Humanity's Last ExamGLM-5.2 by 18.941.122.2+18.9
- GPQA DiamondGLM-5.2 by 5.489.584.1+5.4
FactsGLM-5.2
- AA-Omniscience · Non-hallucinationGLM-5.2 by 24.273.749.5+24.2
- AA-Omniscience · AccuracyGLM-5.2 by 5.524.318.8+5.5
MathGLM-5.2
- FrontierMath Tiers 1-3 (v2)GLM-5.2 by 38.859.220.4+38.8
- IMOAnswerBenchGLM-5.2 by 12.19178.9+12.1
- HMMT Feb. 2026GLM-5.2 by 8.992.583.6+8.9
- AIME 2026GLM-5.2 by 6.599.292.7+6.5
Other results12 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.
- GDPval-AA v2GLM-5.2 by 493 rating points15081015+493 rating
- Artificial Analysis Coding IndexGLM-5.2 by 26.968.841.9+26.9
- AA-OmniscienceGLM-5.2 by 26.64.4-22.2+26.6
- AA Agentic IndexGLM-5.2 by 24.439.415+24.4
- Tool-DecathlonGLM-5.2 by 21.348.226.9+21.3
- ToolathlonGLM-5.2 by 21.348.226.9+21.3
- NL2RepoGLM-5.2 by 19.548.929.4+19.5
- τ³-Bench BankingGLM-5.2 by 16.326.810.5+16.3
- AA IntelligenceGLM-5.2 by 15.533.718.2+15.5
- HMMT 2025GLM-5.2 by 5.394.489.1+5.3
- HMMT Nov. 2025GLM-5.2 by 5.394.489.1+5.3
- τ²-Bench Telecom (AA run)GLM-5.2 by 3.899.195.3+3.8
Questions people ask
Which is better, GLM-5.2 or Qwen3.6 35B A3B?
GLM-5.2 wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. Qwen3.6 35B A3B wins none, but costs 55% less.
Which is better for coding?
GLM-5.2. It wins 4 of the 4 coding tests both models report; Qwen3.6 35B A3B wins none.
Which is cheaper?
GLM-5.2 costs $1.40 per million input tokens and $4.40 per million output tokens; Qwen3.6 35B A3B costs $0.38 and $2.25. That makes Qwen3.6 35B A3B about 55% cheaper for the same work.
How do you compare the two?
We use the 31 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 12 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.