GLM-5.2 vs GPT-5.5
Wins 0 of 7 areas
—
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
GPT-5.5 is the stronger all-rounder.GLM-5.2 is cheaper.
Scores updated · 59 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.
- ReasoningHard problems that need careful thinking06GPT-5.56 of 6 tests
- MathCompetition and research-level math04GPT-5.54 of 5 tests · 1 tie
- CodingWriting and fixing software25GPT-5.55 of 7 tests
- AgentsCarrying out multi-step tasks on its own24GPT-5.54 of 6 tests
- Following instructionsDoing exactly what it is asked02GPT-5.52 of 2 tests
- FactsGetting facts right instead of making them up12GPT-5.52 of 3 tests
- Long documentsFinding answers in very long texts01GPT-5.51 of 1 test
Long documents rests on a single test.
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.
GLM-5.2 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 GPT-5.5 pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+62.2points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+43.2points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+33.7points ahead
Where GLM-5.2 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+62.7points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+3.5points ahead
- Multi-step tasks using many real software toolsMCP Atlas+1.5points ahead
Every test, side by side
All 59 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.5
- Terminal-Bench HardGPT-5.5 by 9.850.860.6+9.8
- LMArena · WebDevGLM-5.2 by 93 rating points16051512+93 rating
- Terminal-Bench 2.1GPT-5.5 by 6.477.984.3+6.4
- SciCodeGPT-5.5 by 4.651.255.8+4.6
- SWE-bench ProGLM-5.2 by 3.562.158.6+3.5
- LiveBench · CodingGPT-5.5 by 2.579.782.2+2.5
- LiveBench · Agentic CodingGPT-5.5 by 2.251.854+2.2
AgentsGPT-5.5
- Terminal-Bench 4.0GPT-5.5 by 13.6114.6+13.6
- τ-Bench V3 · BankingGPT-5.5 by 4.434.639+4.4
- AA ApexAgentsGPT-5.5 by 433.737.7+4
- AA IT-Bench SREGPT-5.5 by 3.142.745.8+3.1
- MCP AtlasGLM-5.2 by 1.576.875.3+1.5
- GDPValGLM-5.2 by 143.742.7+1
ReasoningGPT-5.5
- ARC-AGI-2GPT-5.5 by 62.222.885+62.2
- LiveBench · ReasoningGPT-5.5 by 11.178.689.7+11.1
- SimpleBenchGPT-5.5 by 10.258.869+10.2
- CritPtGPT-5.5 by 6.220.927.1+6.2
- Humanity's Last ExamGPT-5.5 by 4.741.145.8+4.7
- GPQA DiamondGPT-5.5 by 489.593.5+4
FactsGPT-5.5
- AA-Omniscience · Non-hallucinationGLM-5.2 by 62.773.711+62.7
- AA-Omniscience · AccuracyGPT-5.5 by 33.724.358+33.7
- SimpleQA VerifiedGPT-5.5 by 28.834.263+28.8
MathGPT-5.5
- FrontierMath Tier 4GPT-5.5 by 43.229.372.5+43.2
- FrontierMath Tiers 1-3 (v2)GPT-5.5 by 26.159.285.3+26.1
- LiveBench · MathematicsGPT-5.5 by 6.189.895.9+6.1
- HMMT Feb. 2026GPT-5.5 by 692.598.5+6
- AIME 2026tie99.2100tie
Following instructionsGPT-5.5
- LiveBench · Instruction FollowingGPT-5.5 by 8.462.370.7+8.4
- IFBenchGPT-5.5 by 2.673.375.9+2.6
Other results29 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.
- DeepSWEGPT-5.5 by 23.846.270+23.8
- DeepSWE 1.1GPT-5.5 by 234467+23
- ARC-AGI-1GPT-5.5 by 187795+18
- AA-OmniscienceGPT-5.5 by 16.14.420.5+16.1
- Toolathlon VerifiedGPT-5.5 by 13.659.973.5+13.6
- OfficeQA ProGPT-5.5 by 12.741.454.1+12.7
- livebench_languageGPT-5.5 by 11.276.287.4+11.2
- livebench_data_analysisGPT-5.5 by 7.973.781.6+7.9
- Tool-DecathlonGPT-5.5 by 7.448.255.6+7.4
- ToolathlonGPT-5.5 by 7.448.255.6+7.4
- Program BenchGPT-5.5 by 7.163.770.8+7.1
- ResearchRubricsGLM-5.2 by 7.171.164+7.1
- Artificial Analysis Coding IndexGPT-5.5 by 6.168.874.9+6.1
- PostTrainBenchGLM-5.2 by 5.934.328.4+5.9
- τ²-Bench Telecom (AA run)GLM-5.2 by 5.299.193.9+5.2
- JobBenchGLM-5.2 by 5.143.438.3+5.1
- MLS Bench Litelower is betterGPT-5.5 by 4.940.435.5+4.9
- AA IntelligenceGPT-5.5 by 4.733.738.4+4.7
- CyberGymGPT-5.5 by 4.677.281.8+4.6
- HiL-Bench (Tools-allowed)GLM-5.2 by 443.739.7+4
- Agents' Last ExamGPT-5.5 by 2.823.826.6+2.8
- HLE (with tools)GLM-5.2 by 2.554.752.2+2.5
- AA Agentic IndexGLM-5.2 by 2.139.437.3+2.1
- HMMT Nov. 2025GPT-5.5 by 2.194.496.5+2.1
- GDPval-AA v2 EloGLM-5.2 by 19 rating points15101491+19 rating
- NL2RepoGPT-5.5 by 1.848.950.7+1.8
- SpreadsheetBench 2GPT-5.5 by 128.129.1+1
- SWE-MarathonGLM-5.2 by 11312+1
- GDPval-AA v2tie15081509tie
Questions people ask
Which is better, GLM-5.2 or GPT-5.5?
GPT-5.5 wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. GLM-5.2 wins none, but costs 83% less.
Which is better for coding?
GPT-5.5. It wins 5 of the 7 coding tests both models report; GLM-5.2 wins 2.
Which is cheaper?
GLM-5.2 costs $1.40 per million input tokens and $4.40 per million output tokens; GPT-5.5 costs $5.00 and $30.00. That makes GLM-5.2 about 83% cheaper for the same work.
How do you compare the two?
We use the 59 benchmark tests both models have published scores on. The verdict counts the 30 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 29 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.