GLM-5.1 vs GLM-5.2
Wins 1 of 7 areas
Following instructions
Wins 6 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents
GLM-5.2 is the stronger all-rounder.GLM-5.1 is better at following instructions.
Scores updated · 38 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 software05GLM-5.25 of 5 tests
- ReasoningHard problems that need careful thinking04GLM-5.24 of 4 tests
- MathCompetition and research-level math04GLM-5.24 of 4 tests
- AgentsCarrying out multi-step tasks on its own14GLM-5.24 of 5 tests
- FactsGetting facts right instead of making them up01GLM-5.21 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts01GLM-5.21 of 1 test
- Following instructionsDoing exactly what it is asked10GLM-5.11 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.
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- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+22.4points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+21points ahead
- Research-level physics problemsCritPt+16.3points ahead
Where GLM-5.1 pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+3points ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+1points ahead
Every test, side by side
All 38 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 16.161.877.9+16.1
- LMArena · WebDevGLM-5.2 by 95 rating points15081603+95 rating
- Terminal-Bench HardGLM-5.2 by 7.643.250.8+7.6
- SciCodeGLM-5.2 by 6.444.851.2+6.4
- SWE-bench ProGLM-5.2 by 3.758.462.1+3.7
AgentsGLM-5.2
- τ-Bench V3 · BankingGLM-5.2 by 2113.634.6+21
- GDPValGLM-5.2 by 12.73143.7+12.7
- MCP AtlasGLM-5.2 by 571.876.8+5
- AA IT-Bench SREGLM-5.2 by 2.440.342.7+2.4
- Terminal-Bench 4.0GLM-5.1 by 121+1
ReasoningGLM-5.2
- CritPtGLM-5.2 by 16.34.620.9+16.3
- Humanity's Last ExamGLM-5.2 by 1130.141.1+11
- SimpleBenchGLM-5.2 by 3.755.158.8+3.7
- GPQA DiamondGLM-5.2 by 2.786.889.5+2.7
FactsGLM-5.2
- AA-Omniscience · Non-hallucinationGLM-5.2 by 3.670.173.7+3.6
- AA-Omniscience · Accuracytie23.724.3tie
- SimpleQA Verifiedtie3434.2tie
MathGLM-5.2
- FrontierMath Tiers 1-3 (v2)GLM-5.2 by 22.436.859.2+22.4
- HMMT Feb. 2026GLM-5.2 by 9.982.692.5+9.9
- IMOAnswerBenchGLM-5.2 by 7.283.891+7.2
- AIME 2026GLM-5.2 by 3.995.399.2+3.9
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.
- DeepSWEGLM-5.2 by 28.21846.2+28.2
- AA Agentic IndexGLM-5.2 by 14.225.239.4+14.2
- PostTrainBenchGLM-5.2 by 14.220.134.3+14.2
- Artificial Analysis Coding IndexGLM-5.2 by 1355.868.8+13
- Program BenchGLM-5.2 by 12.850.963.7+12.8
- CyberGymGLM-5.2 by 8.568.777.2+8.5
- AA IntelligenceGLM-5.2 by 7.626.133.7+7.6
- Tool-DecathlonGLM-5.2 by 7.540.748.2+7.5
- ToolathlonGLM-5.2 by 7.540.748.2+7.5
- NL2RepoGLM-5.2 by 6.242.748.9+6.2
- AA-OmniscienceGLM-5.2 by 3.50.84.4+3.5
- HLE (with tools)GLM-5.2 by 2.452.354.7+2.4
- τ²-Bench Telecom (AA run)GLM-5.2 by 1.497.799.1+1.4
- HMMT 2025tie9494.4tie
- HMMT Nov. 2025tie9494.4tie
Questions people ask
Which is better, GLM-5.1 or GLM-5.2?
GLM-5.2 wins six of the seven areas where both have results: coding, agents, reasoning, facts, math and long documents. GLM-5.1 wins following instructions.
Which is better for coding?
GLM-5.2. It wins 5 of the 5 coding tests both models report; GLM-5.1 wins none.
Which is cheaper?
GLM-5.1 costs $1.38 per million input tokens and $4.40 per million output tokens; GLM-5.2 costs $1.40 and $4.40. For a million tokens read plus a million written, they cost about the same.
How do you compare the two?
We use the 38 benchmark tests both models have published scores on. The verdict counts the 23 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.