Model comparison
DeepSeek-V3 vs GLM-5.2
GLM-5.2 is the stronger model overall, scoring 51.1 to 39.5 on the Noometry Index. DeepSeek-V3 costs 5.3× less per token, which makes it the better buy when GLM-5.2's lead doesn't matter for your workload.
Last verified . 28 shared benchmarks.
DeepSeekSummary
- They share 28 benchmarks with published results for both. DeepSeek-V3 scores higher in 0 categories and GLM-5.2 in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GLM-5.2 leads 55.7 to 32.1.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 86.4% for GLM-5.2.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $1.40 / $4.40 for GLM-5.2.
- GLM-5.2 accepts more context: 1M tokens versus 164K.
Side by side
| DeepSeek-V3 | GLM-5.2 | |
|---|---|---|
| Provider | DeepSeek | Z.ai (Zhipu) |
| Noometry Index | 39.5 | 51.1 |
| Released | 2024-12-26 | 2026-06-13 |
| Weights | Open | Open |
| Context window | 164K | 1M |
| Max output | 164K | 131K |
| Input $ / M tokens | $0.24 | $1.40 |
| Output $ / M tokens | $0.90 | $4.40 |
| Results tracked | 60 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GLM-5.2 leads
DeepSeek-V3: 42.3 (#106), GLM-5.2: 51.3 (#41)
| Benchmark | DeepSeek-V3 | GLM-5.2 |
|---|---|---|
| SciCode | 35.8% | 50.5% |
| WeirdML | 36.1% | 70.1% |
| LMArena Coding | 1368 | 1485 |
| SWE-bench Verified | — | 78.7% |
| DeepSWE | — | 43.8% |
| FrontierCode | — | 24.5% |
| Aider Polyglot | 55.1% | — |
| LMArena WebDev | — | 1603 |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 1,047 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, GLM-5.2: 32.4 (#63)
| Benchmark | DeepSeek-V3 | GLM-5.2 |
|---|---|---|
| APEX-Agents | — | 45.2% |
| τ²-bench Banking | — | 37.1% |
| PostTrainBench | — | 31.7% |
| GBAEval | — | 0% |
| METR Time Horizons | 49.6% | — |
| Vending-Bench 2 | — | 8,314 |
Reasoning GLM-5.2 leads
DeepSeek-V3: 20.5 (#236), GLM-5.2: 42.3 (#52)
| Benchmark | DeepSeek-V3 | GLM-5.2 |
|---|---|---|
| SimpleBench | 27.2% | 58.8% |
| Kagi LLM Benchmark | 52.3% | 62.6% |
| CritPt | 0% | 20.9% |
| LMArena Hard Prompts | 1365 | 1480 |
| DTBench | 64.8% | 93.6% |
| LMCA | 15.5% | 45.8% |
| Epoch Capabilities Index | 135.94 | 151.78 |
| ARC-AGI-2 | — | 22.8% |
| NYT Connections (extended) | — | 74.3% |
| ARC-AGI-1 | — | 77% |
| Chess Puzzles | — | 21% |
| EBR-Bench | — | 9.5% |
| LiveBench Reasoning | 65.8% | — |
| Mystery Game Puzzles | — | 19% |
| LiveBench Data Analysis | 60.9% | — |
| Surface Evolver Bench | — | 55.6% |
| BIG-Bench Hard | 87.5% | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math GLM-5.2 leads
DeepSeek-V3: 32.1 (#219), GLM-5.2: 55.7 (#43)
| Benchmark | DeepSeek-V3 | GLM-5.2 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 86.4% |
| LMArena Math | 1373 | 1482 |
| FrontierMath (Tiers 1-3) | — | 59.2% |
| FrontierMath Tier 4 | — | 29.3% |
| MathArena Final-Answer Competitions | — | 67.6% |
| ProofBench | — | 35% |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |
Knowledge GLM-5.2 leads
DeepSeek-V3: 37.5 (#155), GLM-5.2: 57.1 (#40)
| Benchmark | DeepSeek-V3 | GLM-5.2 |
|---|---|---|
| GPQA Diamond | 67.6% | 91.9% |
| LMArena Expert | 1351 | 1486 |
| SimpleQA Verified | — | 34.2% |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multilingual GLM-5.2 leads
DeepSeek-V3: 48.5 (#143), GLM-5.2: 55.8 (#26)
| Benchmark | DeepSeek-V3 | GLM-5.2 |
|---|---|---|
| LMArena Non-English | 1358 | 1459 |
| LMArena Chinese | 1391 | 1519 |
| LMArena French | 1385 | 1479 |
| LMArena German | 1374 | 1468 |
| LMArena Japanese | 1333 | 1451 |
| LMArena Korean | 1319 | 1445 |
| LMArena Russian | 1373 | 1466 |
| LMArena Spanish | 1358 | 1477 |
Instruction Following GLM-5.2 leads
DeepSeek-V3: 72.8 (#130), GLM-5.2: 76.9 (#34)
| Benchmark | DeepSeek-V3 | GLM-5.2 |
|---|---|---|
| LMArena Instruction Following | 1345 | 1465 |
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |
Long Context GLM-5.2 leads
DeepSeek-V3: 34.0 (#253), GLM-5.2: 45.3 (#43)
| Benchmark | DeepSeek-V3 | GLM-5.2 |
|---|---|---|
| LMArena Longer Query | 1352 | 1479 |
| Fiction.LiveBench | 50% | — |
Writing & Preference GLM-5.2 leads
DeepSeek-V3: 57.4 (#130), GLM-5.2: 70.4 (#21)
| Benchmark | DeepSeek-V3 | GLM-5.2 |
|---|---|---|
| LMArena Text | 1375 | 1470 |
| LMArena Creative Writing | 1364 | 1462 |
| EQ-Bench Creative Writing | 1472 | 1757 |
| LMArena Multi-Turn | 1389 | 1469 |
| Short-Story Creative Writing | 77% | — |
| WildBench | 83% | — |
| EQ-Bench 4 | — | 1222 |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than GLM-5.2?
GLM-5.2 is the stronger model overall, scoring 51.1 to 39.5 on the Noometry Index. DeepSeek-V3 costs 5.3× less per token, which makes it the better buy when GLM-5.2's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3 or GLM-5.2?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; GLM-5.2 lists at $1.40 and $4.40.
Is DeepSeek-V3 or GLM-5.2 better for coding?
GLM-5.2 scores higher on coding benchmarks: 51.3 versus 42.3 in the Noometry coding category.
Which has the bigger context window?
GLM-5.2 does, with 1M tokens against 164K.
How many benchmarks do DeepSeek-V3 and GLM-5.2 share?
28 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and GLM-5.2 has 51.