Model comparison
DeepSeek V4 Flash vs Inkling
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 44.1 on the Noometry Index.
Last verified . 37 shared benchmarks.
DeepSeekSummary
- They share 37 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 5 categories and Inkling in 3 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Flash leads 60.3 to 31.3.
- The biggest single-benchmark swing is ProofBench: 56% for DeepSeek V4 Flash and 0% for Inkling.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $1.87 / $4.68 for Inkling.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 66K.
Side by side
| DeepSeek V4 Flash | Inkling | |
|---|---|---|
| Provider | DeepSeek | Thinking Machines Lab |
| Noometry Index | 53.6 | 44.1 |
| Released | 2026-04-24 | 2026-07-15 |
| Weights | Open | Open |
| Context window | 1M | 66K |
| Max output | 393K | 66K |
| Input $ / M tokens | $0.15 | $1.87 |
| Output $ / M tokens | $0.60 | $4.68 |
| Results tracked | 41 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Flash leads
DeepSeek V4 Flash: 47.9 (#59), Inkling: 34.5 (#234)
| Benchmark | DeepSeek V4 Flash | Inkling |
|---|---|---|
| FrontierCode | 18.8% | 14% |
| LMArena WebDev | 1582 | 1413 |
| SciCode | 49.9% | 47% |
| WeirdML | 63% | 32.3% |
| LMArena Coding | 1457 | 1464 |
| ALE-Bench | 1,306 | 946 |
| FrontierSWE | — | 4.1% |
Agentic & Tool Use Not comparable
DeepSeek V4 Flash: —, Inkling: 29.6 (#85)
| Benchmark | DeepSeek V4 Flash | Inkling |
|---|---|---|
| APEX-Agents | — | 33.8% |
| τ²-bench Banking | — | 25% |
Reasoning DeepSeek V4 Flash leads
DeepSeek V4 Flash: 53.7 (#30), Inkling: 40.4 (#56)
| Benchmark | DeepSeek V4 Flash | Inkling |
|---|---|---|
| ARC-AGI-2 | 61.4% | 36.5% |
| SimpleBench | 61.1% | 50% |
| ARC-AGI-1 | 89% | 79.5% |
| CritPt | 16.6% | 5.4% |
| Chess Puzzles | 33% | 21% |
| LMArena Hard Prompts | 1444 | 1451 |
| DTBench | 90.9% | 87.5% |
| LMCA | 41.7% | 37.6% |
| Epoch Capabilities Index | 154.49 | 148.54 |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 89.6% | — |
| Mystery Game Puzzles | 34% | — |
Math DeepSeek V4 Flash leads
DeepSeek V4 Flash: 60.3 (#37), Inkling: 31.3 (#225)
| Benchmark | DeepSeek V4 Flash | Inkling |
|---|---|---|
| FrontierMath (Tiers 1-3) | 57.5% | 33.3% |
| FrontierMath Tier 4 | 24.4% | 4.9% |
| OTIS Mock AIME 2024-2025 | 94.4% | 88.9% |
| ProofBench | 56% | 0% |
| LMArena Math | 1427 | 1479 |
| MathArena Final-Answer Competitions | 76.5% | — |
Knowledge Too close to call
DeepSeek V4 Flash: 55.4 (#48), Inkling: 55.1 (#49)
| Benchmark | DeepSeek V4 Flash | Inkling |
|---|---|---|
| GPQA Diamond | 91% | 88.3% |
| SimpleQA Verified | 33.6% | 40.3% |
| LMArena Expert | 1441 | 1465 |
Multilingual Inkling leads
DeepSeek V4 Flash: 53.0 (#72), Inkling: 54.0 (#52)
| Benchmark | DeepSeek V4 Flash | Inkling |
|---|---|---|
| LMArena Non-English | 1420 | 1434 |
| LMArena Chinese | 1468 | 1490 |
| LMArena French | 1439 | 1458 |
| LMArena German | 1418 | 1446 |
| LMArena Japanese | 1406 | 1429 |
| LMArena Korean | 1384 | 1404 |
| LMArena Russian | 1428 | 1429 |
| LMArena Spanish | 1436 | 1448 |
Instruction Following Too close to call
DeepSeek V4 Flash: 74.9 (#81), Inkling: 75.1 (#71)
| Benchmark | DeepSeek V4 Flash | Inkling |
|---|---|---|
| LMArena Instruction Following | 1421 | 1426 |
Long Context Too close to call
DeepSeek V4 Flash: 43.8 (#85), Inkling: 43.8 (#86)
| Benchmark | DeepSeek V4 Flash | Inkling |
|---|---|---|
| LMArena Longer Query | 1434 | 1434 |
Writing & Preference Inkling leads
DeepSeek V4 Flash: 63.8 (#61), Inkling: 65.2 (#51)
| Benchmark | DeepSeek V4 Flash | Inkling |
|---|---|---|
| LMArena Text | 1432 | 1441 |
| LMArena Creative Writing | 1403 | 1387 |
| EQ-Bench Creative Writing | 1559 | 1611 |
| LMArena Multi-Turn | 1449 | 1436 |
| EQ-Bench 4 | — | 1226 |
Frequently asked questions
Is DeepSeek V4 Flash better than Inkling?
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 44.1 on the Noometry Index.
Which is cheaper, DeepSeek V4 Flash or Inkling?
DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; Inkling lists at $1.87 and $4.68.
Is DeepSeek V4 Flash or Inkling better for coding?
DeepSeek V4 Flash scores higher on coding benchmarks: 47.9 versus 34.5 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Flash does, with 1M tokens against 66K.
How many benchmarks do DeepSeek V4 Flash and Inkling share?
37 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and Inkling has 41.