GLM-4.6 vs GLM-4.7
Wins 0 of 7 areas
—
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
GLM-4.7 is the stronger all-rounder.GLM-4.6 is cheaper.
Scores updated · 39 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 software16GLM-4.76 of 7 tests
- ReasoningHard problems that need careful thinking02GLM-4.72 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own12GLM-4.72 of 3 tests
- FactsGetting facts right instead of making them up12GLM-4.72 of 3 tests
- MathCompetition and research-level math01GLM-4.71 of 1 test
- Long documentsFinding answers in very long texts01GLM-4.71 of 1 test
- Following instructionsDoing exactly what it is asked01GLM-4.71 of 1 test
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.
GLM-4.6 costs 1% 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-4.7 pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+24.5points ahead
- Reasons across sets of long documentsAA-LCR+17points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+12.9points ahead
Where GLM-4.6 pulls ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+4.1points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+1.2points ahead
Every test, side by side
All 39 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGLM-4.7
- SWE-bench MultilingualGLM-4.7 by 12.953.866.7+12.9
- LMArena · WebDevGLM-4.7 by 94 rating points13411435+94 rating
- Terminal-Bench HardGLM-4.7 by 6.82531.8+6.8
- SciCodeGLM-4.7 by 6.738.445.1+6.7
- SWE-bench VerifiedGLM-4.7 by 5.86873.8+5.8
- Terminal-Bench 2.1GLM-4.6 by 4.149.445.3+4.1
- LiveCodeBench v6GLM-4.7 by 2.182.884.9+2.1
AgentsGLM-4.7
- GDPValGLM-4.7 by 12.81325.8+12.8
- BrowseCompGLM-4.7 by 6.945.152+6.9
- τ-Bench V3 · BankingGLM-4.6 by 1.213.412.2+1.2
ReasoningGLM-4.7
- Humanity's Last ExamGLM-4.7 by 12.914.527.4+12.9
- GPQA DiamondGLM-4.7 by 7.97885.9+7.9
- CritPttie1.11.7tie
FactsGLM-4.7
- AA-Omniscience · AccuracyGLM-4.7 by 2.426.929.3+2.4
- Vectara HHEM hallucination ratelower is betterGLM-4.6 by 2.29.511.7+2.2
- AA-Omniscience · Non-hallucinationGLM-4.7 by 1.15.97+1.1
Other results20 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.
- τ²-Bench Telecom (AA run)GLM-4.7 by 25.470.595.9+25.4
- xbench-DeepSearchGLM-4.6 by 17.77052.3+17.7
- BrowseComp-ZHGLM-4.7 by 17.149.566.6+17.1
- Terminal-Bench 2.0GLM-4.7 by 16.524.541+16.5
- HLE (with tools)GLM-4.7 by 12.430.442.8+12.4
- τ²-BenchGLM-4.7 by 12.275.287.4+12.2
- browsecomp_with_context_managerGLM-4.7 by 1057.567.5+10
- HMMT Feb. 2025GLM-4.7 by 7.989.297.1+7.9
- AA Agentic IndexGLM-4.7 by 7.618.626.2+7.6
- Terminal-BenchGLM-4.6 by 7.240.533.3+7.2
- vectara_avg_summary_lengthGLM-4.6 by 6.677.270.6+6.6
- AA-OmniscienceGLM-4.7 by 5.5-41.9-36.4+5.5
- vectara_answer_rateGLM-4.7 by 5.394.599.8+5.3
- AA IntelligenceGLM-4.7 by 3.718.522.2+3.7
- vectara_factual_consistencyGLM-4.6 by 2.290.588.3+2.2
- frontiermath_tier_4_v1GLM-4.6 by 2.12.10+2.1
- AIME 2025GLM-4.7 by 1.893.995.7+1.8
- HMMT Nov. 2025GLM-4.7 by 1.891.793.5+1.8
- MMLU-ProGLM-4.7 by 1.183.284.3+1.1
- Artificial Analysis Coding Indextie45.845.3tie
Questions people ask
Which is better, GLM-4.6 or GLM-4.7?
GLM-4.7 wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. GLM-4.6 wins none, but costs 1% less.
Which is better for coding?
GLM-4.7. It wins 6 of the 7 coding tests both models report; GLM-4.6 wins 1.
Which is cheaper?
GLM-4.6 costs $0.57 per million input tokens and $2.20 per million output tokens; GLM-4.7 costs $0.60 and $2.20. That makes GLM-4.6 about 1% cheaper for the same work.
How do you compare the two?
We use the 39 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 20 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.