Model comparison
DeepSeek-V3.2-Exp vs GLM-5
GLM-5 is the stronger model overall, scoring 46.1 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 5.3× less per token, which makes it the better buy when GLM-5's lead doesn't matter for your workload.
Last verified . 37 shared benchmarks.
DeepSeekSummary
- They share 37 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 2 categories and GLM-5 in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GLM-5 leads 27.6 to 22.1.
- The biggest single-benchmark swing is NYT Connections (extended): 36.7% for DeepSeek-V3.2-Exp and 74.8% for GLM-5.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $1 / $3.20 for GLM-5.
- GLM-5 accepts more context: 205K tokens versus 164K.
Side by side
| DeepSeek-V3.2-Exp | GLM-5 | |
|---|---|---|
| Provider | DeepSeek | Z.ai (Zhipu) |
| Noometry Index | 44.3 | 46.1 |
| Released | 2025-09-29 | 2026-02-11 |
| Weights | Open | Open |
| Context window | 164K | 205K |
| Max output | 66K | 131K |
| Input $ / M tokens | $0.26 | $1 |
| Output $ / M tokens | $0.38 | $3.20 |
| Results tracked | 49 | 45 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GLM-5 leads
DeepSeek-V3.2-Exp: 46.5 (#65), GLM-5: 49.0 (#52)
| Benchmark | DeepSeek-V3.2-Exp | GLM-5 |
|---|---|---|
| SWE-bench Verified (bash only) | 70% | 72.8% |
| LMArena WebDev | 1362 | 1434 |
| SWE-bench Multilingual | 59% | 69.7% |
| WeirdML | 39.5% | 48.2% |
| LMArena Coding | 1454 | 1461 |
| SWE-bench Verified | — | 72.1% |
| Aider Polyglot | 74.2% | — |
| SciCode | 38.9% | — |
| ALE-Bench | — | 765.62 |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 32.7 (#59), GLM-5: 31.1 (#71)
| Benchmark | DeepSeek-V3.2-Exp | GLM-5 |
|---|---|---|
| Terminal-Bench | 39.6% | 52.4% |
| Vending-Bench 2 | 1,034 | 4,432 |
| APEX-Agents | 21.3% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| τ²-bench Airline | — | 82.5% |
| τ²-bench Banking | — | 9.8% |
| τ²-bench Retail | — | 73.7% |
| τ²-bench Telecom | — | 86.8% |
Reasoning GLM-5 leads
DeepSeek-V3.2-Exp: 22.1 (#208), GLM-5: 27.6 (#116)
| Benchmark | DeepSeek-V3.2-Exp | GLM-5 |
|---|---|---|
| ARC-AGI-2 | 4% | 4.9% |
| Kagi LLM Benchmark | 52.2% | 75% |
| NYT Connections (extended) | 36.7% | 74.8% |
| ARC-AGI-1 | 57% | 44.7% |
| Chess Puzzles | 14% | 10% |
| LMArena Hard Prompts | 1434 | 1452 |
| Epoch Capabilities Index | 146.27 | 145.83 |
| SimpleBench | — | 53.2% |
| CritPt | 2.9% | — |
| Thematic Generalization | 65% | — |
| DTBench | 87.7% | — |
| LMCA | 29.1% | — |
| ForecastBench | — | 61 |
Math GLM-5 leads
DeepSeek-V3.2-Exp: 41.7 (#87), GLM-5: 46.4 (#71)
| Benchmark | DeepSeek-V3.2-Exp | GLM-5 |
|---|---|---|
| MathArena Final-Answer Competitions | 57.7% | 65.7% |
| OTIS Mock AIME 2024-2025 | 87.8% | 80% |
| LMArena Math | 1435 | 1440 |
| FrontierMath (Feb 2025 set) | 22.1% | 16.4% |
| FrontierMath Tier 4 (v1) | 2.1% | 2.1% |
| ProofBench | 8% | — |
Knowledge Too close to call
DeepSeek-V3.2-Exp: 51.7 (#66), GLM-5: 52.3 (#64)
| Benchmark | DeepSeek-V3.2-Exp | GLM-5 |
|---|---|---|
| GPQA Diamond | 83.4% | 87.8% |
| Vectara Hallucination Rate | 5.3% | 10.1% |
| LMArena Expert | 1436 | 1454 |
Multilingual GLM-5 leads
DeepSeek-V3.2-Exp: 52.2 (#90), GLM-5: 53.7 (#58)
| Benchmark | DeepSeek-V3.2-Exp | GLM-5 |
|---|---|---|
| LMArena Non-English | 1409 | 1430 |
| LMArena Chinese | 1461 | 1511 |
| LMArena French | 1433 | 1455 |
| LMArena German | 1440 | 1445 |
| LMArena Japanese | 1374 | 1416 |
| LMArena Korean | 1371 | 1423 |
| LMArena Russian | 1424 | 1436 |
| LMArena Spanish | 1440 | 1454 |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.5 (#93), GLM-5: 75.2 (#67)
| Benchmark | DeepSeek-V3.2-Exp | GLM-5 |
|---|---|---|
| LMArena Instruction Following | 1413 | 1428 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), GLM-5: 44.7 (#60)
| Benchmark | DeepSeek-V3.2-Exp | GLM-5 |
|---|---|---|
| CL-bench | 13.2% | 18.7% |
| LMArena Longer Query | 1428 | 1446 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference GLM-5 leads
DeepSeek-V3.2-Exp: 62.4 (#77), GLM-5: 66.0 (#38)
| Benchmark | DeepSeek-V3.2-Exp | GLM-5 |
|---|---|---|
| LMArena Text | 1425 | 1446 |
| LMArena Creative Writing | 1403 | 1439 |
| EQ-Bench Creative Writing | 1515 | 1601 |
| LMArena Multi-Turn | 1427 | 1456 |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than GLM-5?
GLM-5 is the stronger model overall, scoring 46.1 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 5.3× less per token, which makes it the better buy when GLM-5's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.2-Exp or GLM-5?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; GLM-5 lists at $1 and $3.20.
Is DeepSeek-V3.2-Exp or GLM-5 better for coding?
GLM-5 scores higher on coding benchmarks: 49.0 versus 46.5 in the Noometry coding category.
Which has the bigger context window?
GLM-5 does, with 205K tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and GLM-5 share?
37 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and GLM-5 has 45.