Model comparison
DeepSeek-V3.2-Exp vs Grok 4.6
Grok 4.6 is the stronger model overall, scoring 56.9 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 10× less per token, which makes it the better buy when Grok 4.6's lead doesn't matter for your workload.
Last verified . 33 shared benchmarks.
DeepSeekSummary
- They share 33 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 2 categories and Grok 4.6 in 7 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Grok 4.6 leads 61.4 to 22.1.
- The biggest single-benchmark swing is ARC-AGI-2: 4% for DeepSeek-V3.2-Exp and 67.1% for Grok 4.6.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $2 / $6 for Grok 4.6.
- Grok 4.6 accepts more context: 500K tokens versus 164K.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.2-Exp | Grok 4.6 | |
|---|---|---|
| Provider | DeepSeek | xAI |
| Noometry Index | 44.3 | 56.9 |
| Released | 2025-09-29 | 2026-08-12 |
| Weights | Open | Proprietary |
| Context window | 164K | 500K |
| Max output | 66K | 500K |
| Input $ / M tokens | $0.26 | $2 |
| Output $ / M tokens | $0.38 | $6 |
| Results tracked | 49 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Grok 4.6 leads
DeepSeek-V3.2-Exp: 46.5 (#65), Grok 4.6: 58.5 (#16)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| LMArena WebDev | 1362 | 1617 |
| SciCode | 38.9% | 56.5% |
| WeirdML | 39.5% | 67.3% |
| LMArena Coding | 1454 | 1465 |
| DeepSWE | — | 67.5% |
| FrontierCode | — | 48% |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| CursorBench | — | 41.4% |
| SWE-bench Multilingual | 59% | — |
| FrontierSWE | — | 25.3% |
| ALE-Bench | — | 1,508 |
Agentic & Tool Use Grok 4.6 leads
DeepSeek-V3.2-Exp: 32.7 (#59), Grok 4.6: 39.4 (#27)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| APEX-Agents | 21.3% | 65.3% |
| Vending-Bench 2 | 1,034 | 9,047 |
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| GDP.pdf | — | 17.2% |
Reasoning Grok 4.6 leads
DeepSeek-V3.2-Exp: 22.1 (#208), Grok 4.6: 61.4 (#20)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| ARC-AGI-2 | 4% | 67.1% |
| NYT Connections (extended) | 36.7% | 80% |
| ARC-AGI-1 | 57% | 87.5% |
| CritPt | 2.9% | 19.7% |
| Chess Puzzles | 14% | 40% |
| LMArena Hard Prompts | 1434 | 1447 |
| DTBench | 87.7% | 97.3% |
| LMCA | 29.1% | 48.5% |
| Epoch Capabilities Index | 146.27 | 156.44 |
| SimpleBench | — | 75.9% |
| Kagi LLM Benchmark | 52.2% | — |
| Thematic Generalization | 65% | — |
| EBR-Bench | — | 30.5% |
| Mystery Game Puzzles | — | 34% |
Math Grok 4.6 leads
DeepSeek-V3.2-Exp: 41.7 (#87), Grok 4.6: 67.0 (#24)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 99.2% |
| ProofBench | 8% | 51% |
| LMArena Math | 1435 | 1423 |
| FrontierMath (Tiers 1-3) | — | 66% |
| FrontierMath Tier 4 | — | 31.7% |
| MathArena Final-Answer Competitions | 57.7% | — |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge Grok 4.6 leads
DeepSeek-V3.2-Exp: 51.7 (#66), Grok 4.6: 63.3 (#20)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| GPQA Diamond | 83.4% | 94% |
| LMArena Expert | 1436 | 1467 |
| SimpleQA Verified | — | 49.3% |
| Vectara Hallucination Rate | 5.3% | — |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, Grok 4.6: 43.6 (#23)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| LMArena Vision | — | 1263 |
| Blueprint-Bench 2 | — | 33.2% |
| Furniture Assembly | — | 40% |
| LMArena Document | — | 1452 |
Multilingual Too close to call
DeepSeek-V3.2-Exp: 52.2 (#90), Grok 4.6: 53.0 (#74)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| LMArena Non-English | 1409 | 1420 |
| LMArena Chinese | 1461 | 1480 |
| LMArena French | 1433 | 1461 |
| LMArena German | 1440 | 1431 |
| LMArena Japanese | 1374 | 1376 |
| LMArena Korean | 1371 | 1397 |
| LMArena Russian | 1424 | 1422 |
| LMArena Spanish | 1440 | 1404 |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.5 (#93), Grok 4.6: 75.4 (#63)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| LMArena Instruction Following | 1413 | 1431 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), Grok 4.6: 44.5 (#66)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| LMArena Longer Query | 1428 | 1454 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference Too close to call
DeepSeek-V3.2-Exp: 62.4 (#77), Grok 4.6: 62.3 (#80)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4.6 |
|---|---|---|
| LMArena Text | 1425 | 1428 |
| LMArena Creative Writing | 1403 | 1428 |
| LMArena Multi-Turn | 1427 | 1425 |
| EQ-Bench Creative Writing | 1515 | — |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Grok 4.6?
Grok 4.6 is the stronger model overall, scoring 56.9 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 10× less per token, which makes it the better buy when Grok 4.6's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.2-Exp or Grok 4.6?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Grok 4.6 lists at $2 and $6.
Is DeepSeek-V3.2-Exp or Grok 4.6 better for coding?
Grok 4.6 scores higher on coding benchmarks: 58.5 versus 46.5 in the Noometry coding category.
Which has the bigger context window?
Grok 4.6 does, with 500K tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and Grok 4.6 share?
33 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and Grok 4.6 has 49.