Model comparison
DeepSeek V4 Pro vs GLM-5.2
DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 50.9 on the Noometry Index.
Last verified . 44 shared benchmarks.
DeepSeekSummary
- They share 44 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 5 categories and GLM-5.2 in 4 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.4 to 42.1.
- The biggest single-benchmark swing is ARC-AGI-2: 61.3% for DeepSeek V4 Pro and 22.8% for GLM-5.2.
- DeepSeek V4 Pro is cheaper at $0.43 / $0.87 per million input/output tokens, against $1.40 / $4.40 for GLM-5.2.
Side by side
| DeepSeek V4 Pro | GLM-5.2 | |
|---|---|---|
| Provider | DeepSeek | Z.ai (Zhipu) |
| Noometry Index | 54.2 | 50.9 |
| Released | 2026-04-24 | 2026-06-13 |
| Weights | Open | Open |
| Context window | 1M | 1M |
| Max output | 393K | 131K |
| Input $ / M tokens | $0.43 | $1.40 |
| Output $ / M tokens | $0.87 | $4.40 |
| Results tracked | 48 | 50 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.2 (#31), GLM-5.2: 50.9 (#37)
| Benchmark | DeepSeek V4 Pro | GLM-5.2 |
|---|---|---|
| SWE-bench Verified | 77.6% | 78.7% |
| FrontierCode | 28.6% | 24.5% |
| LMArena WebDev | 1582 | 1604 |
| SciCode | 51% | 50.5% |
| WeirdML | 66.2% | 70.1% |
| LMArena Coding | 1470 | 1484 |
| ALE-Bench | 1,403 | 1,047 |
| DeepSWE | — | 43.8% |
Agentic & Tool Use DeepSeek V4 Pro leads
DeepSeek V4 Pro: 32.7 (#54), GLM-5.2: 31.6 (#63)
| Benchmark | DeepSeek V4 Pro | GLM-5.2 |
|---|---|---|
| Vending-Bench 2 | 3,285 | 8,314 |
| APEX-Agents | 47.3% | — |
| τ²-bench Banking | — | 37.1% |
| PostTrainBench | — | 31.7% |
| GBAEval | — | 0% |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.4 (#22), GLM-5.2: 42.1 (#46)
| Benchmark | DeepSeek V4 Pro | GLM-5.2 |
|---|---|---|
| ARC-AGI-2 | 61.3% | 22.8% |
| Kagi LLM Benchmark | 53.5% | 62.6% |
| NYT Connections (extended) | 91.3% | 74.3% |
| ARC-AGI-1 | 90.5% | 77% |
| CritPt | 18% | 20.9% |
| Chess Puzzles | 47% | 21% |
| LMArena Hard Prompts | 1461 | 1480 |
| Mystery Game Puzzles | 43% | 19% |
| DTBench | 93.9% | 93.6% |
| LMCA | 45.5% | 45.8% |
| Surface Evolver Bench | 40% | 55.6% |
| Epoch Capabilities Index | 155.39 | 151.76 |
| SimpleBench | — | 58.8% |
| EBR-Bench | — | 9.5% |
| ForecastBench | 56.1 | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.6 (#24), GLM-5.2: 55.4 (#35)
| Benchmark | DeepSeek V4 Pro | GLM-5.2 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 64.6% | 59.2% |
| FrontierMath Tier 4 | 26.8% | 29.3% |
| MathArena Final-Answer Competitions | 76.6% | 67.6% |
| OTIS Mock AIME 2024-2025 | 98.6% | 86.4% |
| ProofBench | 50% | 35% |
| LMArena Math | 1455 | 1482 |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#28), GLM-5.2: 57.1 (#36)
| Benchmark | DeepSeek V4 Pro | GLM-5.2 |
|---|---|---|
| GPQA Diamond | 91.7% | 91.9% |
| SimpleQA Verified | 52.9% | 34.2% |
| LMArena Expert | 1463 | 1484 |
| Vectara Hallucination Rate | 8.6% | — |
Multilingual GLM-5.2 leads
DeepSeek V4 Pro: 54.5 (#44), GLM-5.2: 56.0 (#25)
| Benchmark | DeepSeek V4 Pro | GLM-5.2 |
|---|---|---|
| LMArena Non-English | 1438 | 1459 |
| LMArena Chinese | 1486 | 1518 |
| LMArena French | 1473 | 1476 |
| LMArena German | 1458 | 1470 |
| LMArena Japanese | 1448 | 1451 |
| LMArena Korean | 1452 | 1445 |
| LMArena Russian | 1452 | 1467 |
| LMArena Spanish | 1457 | 1477 |
Instruction Following Too close to call
DeepSeek V4 Pro: 76.3 (#44), GLM-5.2: 77.1 (#32)
| Benchmark | DeepSeek V4 Pro | GLM-5.2 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1465 |
Long Context Too close to call
DeepSeek V4 Pro: 45.0 (#49), GLM-5.2: 45.3 (#43)
| Benchmark | DeepSeek V4 Pro | GLM-5.2 |
|---|---|---|
| LMArena Longer Query | 1459 | 1479 |
| CL-bench Life | 13.5% | — |
Writing & Preference GLM-5.2 leads
DeepSeek V4 Pro: 65.6 (#45), GLM-5.2: 70.7 (#20)
| Benchmark | DeepSeek V4 Pro | GLM-5.2 |
|---|---|---|
| LMArena Text | 1451 | 1470 |
| LMArena Creative Writing | 1446 | 1465 |
| EQ-Bench Creative Writing | 1553 | 1757 |
| EQ-Bench 4 | 1166 | 1222 |
| LMArena Multi-Turn | 1467 | 1468 |
Frequently asked questions
Is DeepSeek V4 Pro better than GLM-5.2?
DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 50.9 on the Noometry Index.
Which is cheaper, DeepSeek V4 Pro or GLM-5.2?
DeepSeek V4 Pro is cheaper. It lists at $0.43 per million input tokens and $0.87 per million output tokens; GLM-5.2 lists at $1.40 and $4.40.
Is DeepSeek V4 Pro or GLM-5.2 better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.2 versus 50.9 in the Noometry coding category.
Which has the bigger context window?
Both accept 1M tokens.
How many benchmarks do DeepSeek V4 Pro and GLM-5.2 share?
44 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and GLM-5.2 has 50.