Model comparison
DeepSeek-V2.5 (Sep 2024) vs Inkling
Inkling is the stronger model overall, scoring 44.1 to 37.6 on the Noometry Index.
Last verified . 17 shared benchmarks.
DeepSeekSummary
- They share 17 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 1 category and Inkling in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Inkling leads 55.1 to 34.8.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Inkling | |
|---|---|---|
| Provider | DeepSeek | Thinking Machines Lab |
| Noometry Index | 37.6 | 44.1 |
| Released | 2024-09-06 | 2026-07-15 |
| Weights | Open | Open |
| Context window | — | 66K |
| Max output | — | 66K |
| Input $ / M tokens | — | $1.87 |
| Output $ / M tokens | — | $4.68 |
| Results tracked | 22 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Inkling leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Inkling: 34.5 (#234)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Inkling |
|---|---|---|
| LMArena Coding | 1309 | 1464 |
| FrontierCode | — | 14% |
| Aider Polyglot | 17.8% | — |
| LMArena WebDev | — | 1413 |
| FrontierSWE | — | 4.1% |
| SciCode | — | 47% |
| WeirdML | — | 32.3% |
| BigCodeBench Instruct | 48.6% | — |
| BigCodeBench Complete | 53.2% | — |
| ALE-Bench | — | 946 |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, Inkling: 29.6 (#85)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Inkling |
|---|---|---|
| APEX-Agents | — | 33.8% |
| τ²-bench Banking | — | 25% |
Reasoning Inkling leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Inkling: 40.4 (#56)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Inkling |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1451 |
| ARC-AGI-2 | — | 36.5% |
| SimpleBench | — | 50% |
| ARC-AGI-1 | — | 79.5% |
| CritPt | — | 5.4% |
| Chess Puzzles | — | 21% |
| DTBench | — | 87.5% |
| LMCA | — | 37.6% |
| Epoch Capabilities Index | — | 148.54 |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Inkling: 31.3 (#225)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Inkling |
|---|---|---|
| LMArena Math | 1288 | 1479 |
| FrontierMath (Tiers 1-3) | — | 33.3% |
| FrontierMath Tier 4 | — | 4.9% |
| OTIS Mock AIME 2024-2025 | — | 88.9% |
| ProofBench | — | 0% |
Knowledge Inkling leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Inkling: 55.1 (#49)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Inkling |
|---|---|---|
| LMArena Expert | 1266 | 1465 |
| GPQA Diamond | — | 88.3% |
| SimpleQA Verified | — | 40.3% |
Multilingual Inkling leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Inkling: 54.0 (#52)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Inkling |
|---|---|---|
| LMArena Non-English | 1273 | 1434 |
| LMArena Chinese | 1318 | 1490 |
| LMArena French | 1289 | 1458 |
| LMArena German | 1258 | 1446 |
| LMArena Japanese | 1228 | 1429 |
| LMArena Korean | 1209 | 1404 |
| LMArena Russian | 1289 | 1429 |
| LMArena Spanish | 1248 | 1448 |
Instruction Following Inkling leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Inkling: 75.1 (#71)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Inkling |
|---|---|---|
| LMArena Instruction Following | 1280 | 1426 |
Long Context Inkling leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Inkling: 43.8 (#86)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Inkling |
|---|---|---|
| LMArena Longer Query | 1301 | 1434 |
Writing & Preference Inkling leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Inkling: 65.2 (#51)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Inkling |
|---|---|---|
| LMArena Text | 1294 | 1441 |
| LMArena Creative Writing | 1285 | 1387 |
| LMArena Multi-Turn | 1297 | 1436 |
| EQ-Bench Creative Writing | — | 1611 |
| EQ-Bench 4 | — | 1226 |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Inkling?
Inkling is the stronger model overall, scoring 44.1 to 37.6 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Inkling better for coding?
Inkling scores higher on coding benchmarks: 34.5 versus 31.7 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Inkling share?
17 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Inkling has 41.