Model comparison
DeepSeek-R1 vs GLM-5
GLM-5 is the stronger model overall, scoring 46.1 to 42.3 on the Noometry Index. DeepSeek-R1 costs 1.7× less per token, which makes it the better buy when GLM-5's lead doesn't matter for your workload.
Last verified . 29 shared benchmarks.
DeepSeekSummary
- They share 29 benchmarks with published results for both. DeepSeek-R1 scores higher in 1 category and GLM-5 in 8 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GLM-5 leads 27.6 to 18.6.
- The biggest single-benchmark swing is ARC-AGI-1: 21.2% for DeepSeek-R1 and 44.7% for GLM-5.
- DeepSeek-R1 is cheaper at $0.50 / $2.15 per million input/output tokens, against $1 / $3.20 for GLM-5.
- GLM-5 accepts more context: 205K tokens versus 164K.
- GLM-5 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-R1 | GLM-5 | |
|---|---|---|
| Provider | DeepSeek | Z.ai (Zhipu) |
| Noometry Index | 42.3 | 46.1 |
| Released | 2025-01-20 | 2026-02-11 |
| Weights | Proprietary | Open |
| Context window | 164K | 205K |
| Max output | 64K | 131K |
| Input $ / M tokens | $0.50 | $1 |
| Output $ / M tokens | $2.15 | $3.20 |
| Results tracked | 52 | 45 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GLM-5 leads
DeepSeek-R1: 46.3 (#68), GLM-5: 49.0 (#52)
| Benchmark | DeepSeek-R1 | GLM-5 |
|---|---|---|
| WeirdML | 41.6% | 48.2% |
| LMArena Coding | 1427 | 1461 |
| ALE-Bench | 804.12 | 765.62 |
| SWE-bench Verified | — | 72.1% |
| SWE-bench Verified (bash only) | — | 72.8% |
| Aider Polyglot | 71.4% | — |
| LMArena WebDev | — | 1434 |
| SWE-bench Multilingual | — | 69.7% |
| SciCode | 35.7% | — |
| LiveBench Coding | 66.7% | — |
| AlgoTune | 1.7 | — |
Agentic & Tool Use Too close to call
DeepSeek-R1: 30.7 (#75), GLM-5: 31.1 (#71)
| Benchmark | DeepSeek-R1 | GLM-5 |
|---|---|---|
| Terminal-Bench | — | 52.4% |
| τ²-bench Airline | — | 82.5% |
| τ²-bench Banking | — | 9.8% |
| τ²-bench Retail | — | 73.7% |
| τ²-bench Telecom | — | 86.8% |
| DeepResearch Bench | 35.1% | — |
| BALROG | 34.9% | — |
| METR Time Horizons | 53.8% | — |
| Vending-Bench 2 | — | 4,432 |
Reasoning GLM-5 leads
DeepSeek-R1: 18.6 (#278), GLM-5: 27.6 (#116)
| Benchmark | DeepSeek-R1 | GLM-5 |
|---|---|---|
| ARC-AGI-2 | 1.3% | 4.9% |
| SimpleBench | 40.8% | 53.2% |
| Kagi LLM Benchmark | 69.4% | 75% |
| ARC-AGI-1 | 21.2% | 44.7% |
| LMArena Hard Prompts | 1416 | 1452 |
| Epoch Capabilities Index | 141.29 | 145.83 |
| ForecastBench | 60 | 61 |
| NYT Connections (extended) | — | 74.8% |
| CritPt | 1.1% | — |
| Chess Puzzles | — | 10% |
| LiveBench Reasoning | 83.2% | — |
| LiveBench Data Analysis | 69.8% | — |
| LiveBench | 71.6% | — |
Math GLM-5 leads
DeepSeek-R1: 43.8 (#79), GLM-5: 46.4 (#71)
| Benchmark | DeepSeek-R1 | GLM-5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.4% | 80% |
| LMArena Math | 1400 | 1440 |
| MathArena Final-Answer Competitions | — | 65.7% |
| Omni-MATH | 42.4% | — |
| LiveBench Math | 80.7% | — |
| MATH Level 5 | 96.6% | — |
| FrontierMath (Feb 2025 set) | — | 16.4% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge GLM-5 leads
DeepSeek-R1: 44.5 (#87), GLM-5: 52.3 (#64)
| Benchmark | DeepSeek-R1 | GLM-5 |
|---|---|---|
| GPQA Diamond | 76.3% | 87.8% |
| Vectara Hallucination Rate | 11.3% | 10.1% |
| LMArena Expert | 1394 | 1454 |
| MMLU-Pro | 79.3% | — |
| Confabulations | 12.7% | — |
| GPQA (HELM) | 66.6% | — |
Multilingual GLM-5 leads
DeepSeek-R1: 52.4 (#85), GLM-5: 53.7 (#58)
| Benchmark | DeepSeek-R1 | GLM-5 |
|---|---|---|
| LMArena Non-English | 1412 | 1430 |
| LMArena Chinese | 1442 | 1511 |
| LMArena French | 1417 | 1455 |
| LMArena German | 1404 | 1445 |
| LMArena Japanese | 1391 | 1416 |
| LMArena Korean | 1360 | 1423 |
| LMArena Russian | 1423 | 1436 |
| LMArena Spanish | 1411 | 1454 |
Instruction Following GLM-5 leads
DeepSeek-R1: 72.0 (#143), GLM-5: 75.2 (#67)
| Benchmark | DeepSeek-R1 | GLM-5 |
|---|---|---|
| LMArena Instruction Following | 1382 | 1428 |
| LiveBench Instruction Following | 80.5% | — |
| IFEval | 78.4% | — |
Long Context Too close to call
DeepSeek-R1: 45.4 (#36), GLM-5: 44.7 (#60)
| Benchmark | DeepSeek-R1 | GLM-5 |
|---|---|---|
| LMArena Longer Query | 1391 | 1446 |
| Fiction.LiveBench | 75% | — |
| CL-bench | — | 18.7% |
Writing & Preference GLM-5 leads
DeepSeek-R1: 61.4 (#88), GLM-5: 66.0 (#38)
| Benchmark | DeepSeek-R1 | GLM-5 |
|---|---|---|
| LMArena Text | 1428 | 1446 |
| LMArena Creative Writing | 1405 | 1439 |
| EQ-Bench Creative Writing | 1500 | 1601 |
| LMArena Multi-Turn | 1405 | 1456 |
| Short-Story Creative Writing | 83% | — |
| WildBench | 82.8% | — |
| LiveBench Language | 48.5% | — |
Frequently asked questions
Is DeepSeek-R1 better than GLM-5?
GLM-5 is the stronger model overall, scoring 46.1 to 42.3 on the Noometry Index. DeepSeek-R1 costs 1.7× less per token, which makes it the better buy when GLM-5's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-R1 or GLM-5?
DeepSeek-R1 is cheaper. It lists at $0.50 per million input tokens and $2.15 per million output tokens; GLM-5 lists at $1 and $3.20.
Is DeepSeek-R1 or GLM-5 better for coding?
GLM-5 scores higher on coding benchmarks: 49.0 versus 46.3 in the Noometry coding category.
Which has the bigger context window?
GLM-5 does, with 205K tokens against 164K.
How many benchmarks do DeepSeek-R1 and GLM-5 share?
29 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and GLM-5 has 45.