Model comparison
DeepSeek V4 Flash vs GLM-5
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 45.9 on the Noometry Index.
Last verified . 31 shared benchmarks.
DeepSeekSummary
- They share 31 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 3 categories and GLM-5 in 5 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Flash leads 53.7 to 27.6.
- The biggest single-benchmark swing is ARC-AGI-2: 61.4% for DeepSeek V4 Flash and 4.9% for GLM-5.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $1 / $3.20 for GLM-5.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 205K.
Side by side
| DeepSeek V4 Flash | GLM-5 | |
|---|---|---|
| Provider | DeepSeek | Z.ai (Zhipu) |
| Noometry Index | 53.6 | 45.9 |
| Released | 2026-04-24 | 2026-02-11 |
| Weights | Open | Open |
| Context window | 1M | 205K |
| Max output | 393K | 131K |
| Input $ / M tokens | $0.15 | $1 |
| Output $ / M tokens | $0.60 | $3.20 |
| Results tracked | 41 | 45 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GLM-5 leads
DeepSeek V4 Flash: 47.6 (#56), GLM-5: 49.0 (#47)
| Benchmark | DeepSeek V4 Flash | GLM-5 |
|---|---|---|
| LMArena WebDev | 1581 | 1434 |
| WeirdML | 63% | 48.2% |
| LMArena Coding | 1457 | 1461 |
| ALE-Bench | 1,306 | 765.62 |
| SWE-bench Verified | — | 72.1% |
| FrontierCode | 18.8% | — |
| SWE-bench Verified (bash only) | — | 72.8% |
| SWE-bench Multilingual | — | 69.7% |
| SciCode | 49.9% | — |
Agentic & Tool Use Not comparable
DeepSeek V4 Flash: —, GLM-5: 30.9 (#65)
| Benchmark | DeepSeek V4 Flash | 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% |
| Vending-Bench 2 | — | 4,432 |
Reasoning DeepSeek V4 Flash leads
DeepSeek V4 Flash: 53.7 (#25), GLM-5: 27.6 (#100)
| Benchmark | DeepSeek V4 Flash | GLM-5 |
|---|---|---|
| ARC-AGI-2 | 61.4% | 4.9% |
| SimpleBench | 61.1% | 53.2% |
| Kagi LLM Benchmark | 52.2% | 75% |
| NYT Connections (extended) | 89.6% | 74.8% |
| ARC-AGI-1 | 89% | 44.7% |
| Chess Puzzles | 33% | 10% |
| LMArena Hard Prompts | 1443 | 1451 |
| Epoch Capabilities Index | 154.51 | 145.84 |
| CritPt | 16.6% | — |
| Mystery Game Puzzles | 34% | — |
| DTBench | 90.9% | — |
| LMCA | 41.7% | — |
| ForecastBench | — | 61 |
Math DeepSeek V4 Flash leads
DeepSeek V4 Flash: 60.0 (#31), GLM-5: 44.8 (#66)
| Benchmark | DeepSeek V4 Flash | GLM-5 |
|---|---|---|
| MathArena Final-Answer Competitions | 76.5% | 65.7% |
| OTIS Mock AIME 2024-2025 | 94.4% | 80% |
| LMArena Math | 1427 | 1440 |
| FrontierMath (Tiers 1-3) | 57.5% | — |
| FrontierMath Tier 4 | 24.4% | — |
| ProofBench | 56% | — |
| FrontierMath (Feb 2025 set) | — | 16.4% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge DeepSeek V4 Flash leads
DeepSeek V4 Flash: 55.5 (#43), GLM-5: 52.2 (#59)
| Benchmark | DeepSeek V4 Flash | GLM-5 |
|---|---|---|
| GPQA Diamond | 91% | 87.8% |
| LMArena Expert | 1441 | 1454 |
| SimpleQA Verified | 33.6% | — |
| Vectara Hallucination Rate | — | 10.1% |
Multilingual Too close to call
DeepSeek V4 Flash: 53.2 (#71), GLM-5: 53.9 (#57)
| Benchmark | DeepSeek V4 Flash | GLM-5 |
|---|---|---|
| LMArena Non-English | 1420 | 1430 |
| LMArena Chinese | 1467 | 1511 |
| LMArena French | 1438 | 1455 |
| LMArena German | 1418 | 1445 |
| LMArena Japanese | 1403 | 1416 |
| LMArena Korean | 1383 | 1421 |
| LMArena Russian | 1428 | 1436 |
| LMArena Spanish | 1437 | 1452 |
Instruction Following Too close to call
DeepSeek V4 Flash: 75.1 (#76), GLM-5: 75.4 (#64)
| Benchmark | DeepSeek V4 Flash | GLM-5 |
|---|---|---|
| LMArena Instruction Following | 1421 | 1427 |
Long Context Too close to call
DeepSeek V4 Flash: 43.8 (#81), GLM-5: 44.6 (#58)
| Benchmark | DeepSeek V4 Flash | GLM-5 |
|---|---|---|
| LMArena Longer Query | 1435 | 1446 |
| CL-bench | — | 18.7% |
Writing & Preference GLM-5 leads
DeepSeek V4 Flash: 64.0 (#59), GLM-5: 66.2 (#38)
| Benchmark | DeepSeek V4 Flash | GLM-5 |
|---|---|---|
| LMArena Text | 1432 | 1446 |
| LMArena Creative Writing | 1403 | 1438 |
| EQ-Bench Creative Writing | 1559 | 1601 |
| LMArena Multi-Turn | 1449 | 1456 |
Frequently asked questions
Is DeepSeek V4 Flash better than GLM-5?
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 45.9 on the Noometry Index.
Which is cheaper, DeepSeek V4 Flash or GLM-5?
DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; GLM-5 lists at $1 and $3.20.
Is DeepSeek V4 Flash or GLM-5 better for coding?
GLM-5 scores higher on coding benchmarks: 49.0 versus 47.6 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Flash does, with 1M tokens against 205K.
How many benchmarks do DeepSeek V4 Flash and GLM-5 share?
31 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and GLM-5 has 45.