Model comparison
DeepSeek V4 Pro vs GLM-4.6
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 41.4 on the Noometry Index.
Last verified . 24 shared benchmarks.
DeepSeekSummary
- They share 24 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 9 categories and GLM-4.6 in 0 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 23.7.
- The biggest single-benchmark swing is CritPt: 18% for DeepSeek V4 Pro and 1.1% for GLM-4.6.
- Both cost about the same: $0.66 input and $1.98 output per million tokens.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 205K.
Side by side
| DeepSeek V4 Pro | GLM-4.6 | |
|---|---|---|
| Provider | DeepSeek | Z.ai (Zhipu) |
| Noometry Index | 54.3 | 41.4 |
| Released | 2026-04-24 | 2025-09-30 |
| Weights | Open | Open |
| Context window | 1M | 205K |
| Max output | 393K | 131K |
| Input $ / M tokens | $0.66 | $0.60 |
| Output $ / M tokens | $1.98 | $2.20 |
| Results tracked | 48 | 29 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), GLM-4.6: 40.1 (#148)
| Benchmark | DeepSeek V4 Pro | GLM-4.6 |
|---|---|---|
| LMArena WebDev | 1582 | 1340 |
| SciCode | 51% | 38.4% |
| LMArena Coding | 1470 | 1449 |
| ALE-Bench | 1,403 | 340.82 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| SWE-bench Verified (bash only) | — | 55.4% |
| WeirdML | 66.2% | — |
Agentic & Tool Use Too close to call
DeepSeek V4 Pro: 32.8 (#58), GLM-4.6: 32.3 (#66)
| Benchmark | DeepSeek V4 Pro | GLM-4.6 |
|---|---|---|
| Terminal-Bench | — | 24.5% |
| APEX-Agents | 47.3% | — |
| Berkeley Function Calling Leaderboard | — | 72.4% |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), GLM-4.6: 23.7 (#172)
| Benchmark | DeepSeek V4 Pro | GLM-4.6 |
|---|---|---|
| Kagi LLM Benchmark | 53.5% | 47.4% |
| CritPt | 18% | 1.1% |
| LMArena Hard Prompts | 1461 | 1440 |
| ARC-AGI-2 | 61.3% | — |
| NYT Connections (extended) | 91.3% | — |
| ARC-AGI-1 | 90.5% | — |
| Chess Puzzles | 47% | — |
| Mystery Game Puzzles | 43% | — |
| DTBench | 93.9% | — |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
| Epoch Capabilities Index | 155.31 | — |
| ForecastBench | 56.1 | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), GLM-4.6: 39.1 (#111)
| Benchmark | DeepSeek V4 Pro | GLM-4.6 |
|---|---|---|
| LMArena Math | 1455 | 1432 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| OTIS Mock AIME 2024-2025 | 98.6% | — |
| ProofBench | 50% | — |
| FrontierMath (Feb 2025 set) | — | 3.8% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), GLM-4.6: 40.2 (#124)
| Benchmark | DeepSeek V4 Pro | GLM-4.6 |
|---|---|---|
| Vectara Hallucination Rate | 8.6% | 9.5% |
| LMArena Expert | 1464 | 1431 |
| GPQA Diamond | 91.7% | — |
| SimpleQA Verified | 52.9% | — |
Multilingual Too close to call
DeepSeek V4 Pro: 54.4 (#45), GLM-4.6: 53.5 (#66)
| Benchmark | DeepSeek V4 Pro | GLM-4.6 |
|---|---|---|
| LMArena Non-English | 1439 | 1426 |
| LMArena Chinese | 1486 | 1499 |
| LMArena French | 1472 | 1459 |
| LMArena German | 1458 | 1447 |
| LMArena Japanese | 1445 | 1393 |
| LMArena Korean | 1447 | 1400 |
| LMArena Russian | 1453 | 1419 |
| LMArena Spanish | 1458 | 1436 |
Instruction Following DeepSeek V4 Pro leads
DeepSeek V4 Pro: 76.1 (#47), GLM-4.6: 74.3 (#98)
| Benchmark | DeepSeek V4 Pro | GLM-4.6 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1410 |
Long Context DeepSeek V4 Pro leads
DeepSeek V4 Pro: 45.0 (#51), GLM-4.6: 43.4 (#94)
| Benchmark | DeepSeek V4 Pro | GLM-4.6 |
|---|---|---|
| LMArena Longer Query | 1458 | 1422 |
| CL-bench Life | 13.5% | — |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Pro: 65.5 (#46), GLM-4.6: 61.1 (#90)
| Benchmark | DeepSeek V4 Pro | GLM-4.6 |
|---|---|---|
| LMArena Text | 1451 | 1440 |
| LMArena Creative Writing | 1446 | 1411 |
| EQ-Bench Creative Writing | 1553 | 1411 |
| LMArena Multi-Turn | 1467 | 1427 |
| EQ-Bench 4 | 1166 | — |
Frequently asked questions
Is DeepSeek V4 Pro better than GLM-4.6?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 41.4 on the Noometry Index.
Which is cheaper, DeepSeek V4 Pro or GLM-4.6?
DeepSeek V4 Pro is cheaper. It lists at $0.66 per million input tokens and $1.98 per million output tokens; GLM-4.6 lists at $0.60 and $2.20.
Is DeepSeek V4 Pro or GLM-4.6 better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 40.1 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Pro does, with 1M tokens against 205K.
How many benchmarks do DeepSeek V4 Pro and GLM-4.6 share?
24 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and GLM-4.6 has 29.