DeepSeek-V3.2 vs GLM-4.7
Wins 2 of 7 areas
Facts · Long documents
Wins 3 of 7 areas
Reasoning · Math · Following instructions
GLM-4.7 wins more areas, narrowly.DeepSeek-V3.2 is cheaper and better at facts.
Scores updated · 45 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 thinking12GLM-4.72 of 3 tests
- MathCompetition and research-level math01GLM-4.71 of 1 test
- Following instructionsDoing exactly what it is asked01GLM-4.71 of 1 test
- FactsGetting facts right instead of making them up31DeepSeek-V3.23 of 4 tests
- Long documentsFinding answers in very long texts10DeepSeek-V3.21 of 1 test
- CodingWriting and fixing software33Even3 each · 1 tie
- AgentsCarrying out multi-step tasks on its own11Even1 each · 1 tie
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.
DeepSeek-V3.2 costs 75% 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-4.7 pulls ahead
Where DeepSeek-V3.2 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+10.3points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+6.6points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+3.8points ahead
Every test, side by side
All 45 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- LMArena · WebDevGLM-4.7 by 73 rating points13621435+73 rating
- SciCodeGLM-4.7 by 6.238.945.1+6.2
- Terminal-Bench HardDeepSeek-V3.2 by 3.835.631.8+3.8
- SWE-bench MultilingualDeepSeek-V3.2 by 3.570.266.7+3.5
- LiveCodeBench v6GLM-4.7 by 1.683.384.9+1.6
- Terminal-Bench 2.1DeepSeek-V3.2 by 1.546.845.3+1.5
- SWE-bench Verifiedtie73.173.8tie
AgentsEven
- GDPValGLM-4.7 by 169.825.8+16
- τ-Bench V3 · BankingDeepSeek-V3.2 by 6.618.812.2+6.6
- BrowseComptie51.452tie
ReasoningGLM-4.7
- Humanity's Last ExamGLM-4.7 by 2.824.627.4+2.8
- GPQA DiamondGLM-4.7 by 1.98485.9+1.9
- CritPtDeepSeek-V3.2 by 1.22.91.7+1.2
FactsDeepSeek-V3.2
- AA-Omniscience · Non-hallucinationDeepSeek-V3.2 by 10.317.37+10.3
- Vectara HHEM hallucination ratelower is betterDeepSeek-V3.2 by 5.46.311.7+5.4
- SimpleQA VerifiedGLM-4.7 by 4.727.532.2+4.7
- AA-Omniscience · AccuracyDeepSeek-V3.2 by 3.73329.3+3.7
Other results25 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.
- AA-OmniscienceDeepSeek-V3.2 by 13.9-22.5-36.4+13.9
- GAIA (no file)DeepSeek-V3.2 by 13.275.161.9+13.2
- vectara_avg_summary_lengthGLM-4.7 by 8.66270.6+8.6
- AA Agentic IndexGLM-4.7 by 7.918.326.2+7.9
- vectara_answer_rateGLM-4.7 by 7.292.699.8+7.2
- τ²-BenchGLM-4.7 by 7.180.387.4+7.1
- CyberGymGLM-4.7 by 6.217.323.5+6.2
- ResearchRubricsGLM-4.7 by 6.255.862+6.2
- Terminal-Bench 2.0DeepSeek-V3.2 by 5.446.441+5.4
- vectara_factual_consistencyDeepSeek-V3.2 by 5.493.788.3+5.4
- τ²-Bench Telecom (AA run)GLM-4.7 by 5.390.695.9+5.3
- HMMT Feb. 2025GLM-4.7 by 4.692.597.1+4.6
- Terminal-BenchDeepSeek-V3.2 by 4.437.733.3+4.4
- HMMT Nov. 2025GLM-4.7 by 3.59093.5+3.5
- xbench-DeepSearchDeepSeek-V3.2 by 3.455.752.3+3.4
- xbench-DeepSearch (2025.10)DeepSeek-V3.2 by 3.455.752.3+3.4
- AIME 2025GLM-4.7 by 2.693.195.7+2.6
- frontiermath_tier_4_v1DeepSeek-V3.2 by 2.12.10+2.1
- HLE (with tools)GLM-4.7 by 240.842.8+2
- BrowseComp-ZHGLM-4.7 by 1.66566.6+1.6
- Artificial Analysis Coding IndexGLM-4.7 by 1.144.245.3+1.1
- AA Intelligencetie21.522.2tie
- MMLU-Protie8584.3tie
- AIME 2026 I (Tools-allowed)tie92.792.9tie
- browsecomp_with_context_managertie67.667.5tie
Questions people ask
Which is better, DeepSeek-V3.2 or GLM-4.7?
GLM-4.7 wins three of the seven areas where both have results: reasoning, math and following instructions. DeepSeek-V3.2 wins facts and long documents, and costs 75% less. They are level on coding and agents.
Which is better for coding?
Neither. They win 3 coding tests each of the 7 both models report, and 1 is a tie.
Which is cheaper?
DeepSeek-V3.2 costs $0.28 per million input tokens and $0.42 per million output tokens; GLM-4.7 costs $0.60 and $2.20. That makes DeepSeek-V3.2 about 75% cheaper for the same work.
How do you compare the two?
We use the 45 benchmark tests both models have published scores on. The verdict counts the 20 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 25 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.