GLM-5.3 vs Mistral Large 4
Wins 6 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Following instructions
Wins 1 of 7 areas
Long documents
GLM-5.3 is the stronger all-rounder.Mistral Large 4 is cheaper and better at long documents.
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 software41GLM-5.34 of 5 tests
- ReasoningHard problems that need careful thinking30GLM-5.33 of 3 tests
- FactsGetting facts right instead of making them up30GLM-5.33 of 3 tests
- AgentsCarrying out multi-step tasks on its own20GLM-5.32 of 2 tests
- MathCompetition and research-level math21GLM-5.32 of 3 tests
- Following instructionsDoing exactly what it is asked10GLM-5.31 of 1 test
- Long documentsFinding answers in very long texts01Mistral Large 41 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.
Mistral Large 4 costs 45% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GLM-5.3 pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+18.6points ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+17.2points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+12.3points ahead
Where Mistral Large 4 pulls ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+6.3points ahead
- Fresh competition math problemsLiveBench · Mathematics+5.7points ahead
- Reasons across sets of long documentsAA-LCR+1.6points 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-5.3
- Terminal-Bench 2.1Mistral Large 4 by 6.383.990.2+6.3
- LMArena · WebDevGLM-5.3 by 52 rating points16221570+52 rating
- SciCodeGLM-5.3 by 4.85954.2+4.8
- LiveBench · Agentic CodingGLM-5.3 by 3.760.957.2+3.7
- LiveBench · CodingGLM-5.3 by 1.87977.2+1.8
AgentsGLM-5.3
- Terminal-Bench 4.0GLM-5.3 by 17.241.924.7+17.2
- GDPValGLM-5.3 by 11.357.546.2+11.3
ReasoningGLM-5.3
- CritPtGLM-5.3 by 9.119.110+9.1
- Humanity's Last ExamGLM-5.3 by 7.342.335+7.3
- LiveBench · ReasoningGLM-5.3 by 1.985.883.9+1.9
FactsGLM-5.3
- SimpleQA VerifiedGLM-5.3 by 18.64122.4+18.6
- AA-Omniscience · Non-hallucinationGLM-5.3 by 12.370.458.1+12.3
- AA-Omniscience · AccuracyGLM-5.3 by 8.133.925.8+8.1
MathGLM-5.3
- FrontierMath Tiers 1-3 (v2)GLM-5.3 by 10.268.858.6+10.2
- LiveBench · MathematicsMistral Large 4 by 5.787.993.6+5.7
- FrontierMath Tier 4GLM-5.3 by 2.529.326.8+2.5
Following instructionsGLM-5.3
- LiveBench · Instruction FollowingGLM-5.3 by 4.569.364.8+4.5
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.
- CyberGym-E2E-AAMistral Large 4 by 502979+50
- livebench_languageGLM-5.3 by 30.379.949.6+30.3
- AA-OmniscienceGLM-5.3 by 19.614.3-5.3+19.6
- JailbreakBench (cyber subset): safe-response rateMistral Large 4 by 16.281.197.3+16.2
- AgentHarm (cyber subset): 100 minus harm scoreMistral Large 4 by 7.892.2100+7.8
- CritPt v2 (Hard Subset)GLM-5.3 by 7.533.526+7.5
- CritPt v2 HardGLM-5.3 by 6.533.527+6.5
- AA IntelligenceGLM-5.3 by 6.444.838.4+6.4
- GDP (Surge AI)Mistral Large 4 by 6.411.217.6+6.4
- livebench_data_analysisMistral Large 4 by 6.370.276.5+6.3
- StrongREJECT (cyber subset): refusal rateMistral Large 4 by 5.68792.6+5.6
- DeepSWE 1.1GLM-5.3 by 5.266.961.7+5.2
- Terminal-Bench-Science 0.1GLM-5.3 by 5.29.54.3+5.2
- BrowseComp subsetMistral Large 4 by 2.582.585+2.5
- BrowseComp subset (internal)Mistral Large 4 by 2.582.585+2.5
- CyberGymGLM-5.3 by 2.584.582+2.5
- Terminal-Bench 4.0 (AA Coding Agent Index)Mistral Large 4 by 14041+1
- Terminal-Bench 4.0tie41.841tie
- HLE-Diamondtie17.917.3tie
- BeyondAIMEtie90.690.6tie
Questions people ask
Which is better, GLM-5.3 or Mistral Large 4?
GLM-5.3 wins six of the seven areas where both have results: coding, agents, reasoning, facts, math and following instructions. Mistral Large 4 wins long documents, and costs 45% less.
Which is better for coding?
GLM-5.3. It wins 4 of the 5 coding tests both models report; Mistral Large 4 wins 1.
Which is cheaper?
GLM-5.3 costs $1.13 per million input tokens and $3.96 per million output tokens; Mistral Large 4 costs $0.70 and $2.10. That makes Mistral Large 4 about 45% cheaper for the same work.
How do you compare the two?
We use the 38 benchmark tests both models have published scores on. The verdict counts the 18 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.