Model comparison
Inkling-Small vs Qwen3 235B-A22B
Inkling-Small is the stronger model overall, scoring 46.5 to 43.5 on the Noometry Index.
Last verified . 28 shared benchmarks.
Alibaba (Qwen)Summary
- They share 28 benchmarks with published results for both. Inkling-Small scores higher in 3 categories and Qwen3 235B-A22B in 5 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Inkling-Small leads 38.6 to 15.7.
- The biggest single-benchmark swing is ARC-AGI-1: 84% for Inkling-Small and 11% for Qwen3 235B-A22B.
- Inkling-Small is cheaper at $0.45 / $1.20 per million input/output tokens, against $0.70 / $2.80 for Qwen3 235B-A22B.
- Inkling-Small accepts more context: 524K tokens versus 131K.
Side by side
| Inkling-Small | Qwen3 235B-A22B | |
|---|---|---|
| Provider | Thinking Machines Lab | Alibaba (Qwen) |
| Noometry Index | 46.5 | 43.5 |
| Released | 2026-07-15 | 2025-04 |
| Weights | Open | Open |
| Context window | 524K | 131K |
| Max output | 1.05M | 16K |
| Input $ / M tokens | $0.45 | $0.70 |
| Output $ / M tokens | $1.20 | $2.80 |
| Results tracked | 33 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Inkling-Small: 43.6 (#85), Qwen3 235B-A22B: 44.3 (#75)
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| SciCode | 48.7% | 42.4% |
| LMArena Coding | 1451 | 1445 |
| Aider Polyglot | — | 59.6% |
| LMArena WebDev | 1409 | — |
| WeirdML | — | 41% |
Agentic & Tool Use Not comparable
Inkling-Small: —, Qwen3 235B-A22B: 33.9 (#51)
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 52.1% |
| Vending-Bench 2 | — | -11.34 |
Reasoning Inkling-Small leads
Inkling-Small: 38.6 (#63), Qwen3 235B-A22B: 15.7 (#311)
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| ARC-AGI-2 | 40.1% | 1.3% |
| ARC-AGI-1 | 84% | 11% |
| CritPt | 8.3% | 0% |
| Chess Puzzles | 18% | 12% |
| LMArena Hard Prompts | 1423 | 1433 |
| Mystery Game Puzzles | 6% | 9% |
| Epoch Capabilities Index | 150.15 | 143.85 |
| SimpleBench | — | 31% |
| Kagi LLM Benchmark | — | 69.4% |
| DTBench | — | 80.3% |
| LMCA | — | 29.3% |
| ForecastBench | — | 59.7 |
Math Qwen3 235B-A22B leads
Inkling-Small: 45.1 (#77), Qwen3 235B-A22B: 50.4 (#57)
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 90% | 86.7% |
| LMArena Math | 1459 | 1432 |
| FrontierMath (Tiers 1-3) | 46.3% | — |
| FrontierMath Tier 4 | 17.1% | — |
| ProofBench | 6% | — |
| Omni-MATH | — | 71.8% |
| MATH Level 5 | — | 68.9% |
| FrontierMath (Feb 2025 set) | — | 8.5% |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge Qwen3 235B-A22B leads
Inkling-Small: 48.2 (#77), Qwen3 235B-A22B: 49.6 (#73)
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| GPQA Diamond | 88.5% | 80.1% |
| SimpleQA Verified | 19.1% | 40.4% |
| LMArena Expert | 1442 | 1463 |
| MMLU-Pro | — | 84.4% |
| Confabulations | — | 15.6% |
| Vectara Hallucination Rate | — | 9.3% |
| GPQA (HELM) | — | 72.7% |
Multimodal Not comparable
Inkling-Small: 39.1 (#62), Qwen3 235B-A22B: —
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| LMArena Vision | 1235 | — |
Multilingual Too close to call
Inkling-Small: 51.7 (#104), Qwen3 235B-A22B: 52.3 (#89)
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| LMArena Non-English | 1402 | 1409 |
| LMArena Chinese | 1465 | 1481 |
| LMArena French | 1436 | 1445 |
| LMArena German | 1405 | 1433 |
| LMArena Japanese | 1405 | 1399 |
| LMArena Korean | 1363 | 1391 |
| LMArena Russian | 1391 | 1411 |
| LMArena Spanish | 1428 | 1430 |
Instruction Following Inkling-Small leads
Inkling-Small: 73.8 (#114), Qwen3 235B-A22B: 72.6 (#136)
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| LMArena Instruction Following | 1399 | 1408 |
| IFEval | — | 83.5% |
Long Context Qwen3 235B-A22B leads
Inkling-Small: 42.7 (#118), Qwen3 235B-A22B: 46.1 (#26)
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| LMArena Longer Query | 1401 | 1426 |
| Fiction.LiveBench | — | 75% |
Writing & Preference Too close to call
Inkling-Small: 59.6 (#107), Qwen3 235B-A22B: 59.6 (#108)
| Benchmark | Inkling-Small | Qwen3 235B-A22B |
|---|---|---|
| LMArena Text | 1414 | 1419 |
| LMArena Creative Writing | 1331 | 1384 |
| EQ-Bench Creative Writing | 1491 | 1366 |
| LMArena Multi-Turn | 1418 | 1432 |
| Short-Story Creative Writing | — | 83% |
| WildBench | — | 86.6% |
Frequently asked questions
Is Inkling-Small better than Qwen3 235B-A22B?
Inkling-Small is the stronger model overall, scoring 46.5 to 43.5 on the Noometry Index.
Which is cheaper, Inkling-Small or Qwen3 235B-A22B?
Inkling-Small is cheaper. It lists at $0.45 per million input tokens and $1.20 per million output tokens; Qwen3 235B-A22B lists at $0.70 and $2.80.
Is Inkling-Small or Qwen3 235B-A22B better for coding?
They score almost the same on coding (43.6 vs 44.3); test both on your own repository before choosing.
Which has the bigger context window?
Inkling-Small does, with 524K tokens against 131K.
How many benchmarks do Inkling-Small and Qwen3 235B-A22B share?
28 benchmarks have published results for both models. Inkling-Small has 33 scored results on Noometry and Qwen3 235B-A22B has 49.