Inkling 2 Small vs Qwen4 Max
Full profiles: Inkling 2 Small (Thinking Machines) and Qwen4 Max (Alibaba): every benchmark, price and rank.
Updated · Data from 10 sources
As of 19 Nov 2026, there are not enough shared benchmarks for a verdict, and Inkling 2 Small is 4.7× cheaper at $0.475 vs $2.25 per 1M tokens blended. Choose Inkling 2 Small for coding (59 vs 55) and fast first answers (40.0 s vs 139.2 s), where it also streams ~2× faster; choose Qwen4 Max for reasoning (59 vs 52) and knowledge (59 vs 53).
Key facts (as of 19 Nov 2026)
- Inkling 2 Small costs $0.475 per 1M tokens blended (list price, 19 Nov 2026).
- Qwen4 Max costs $2.25 per 1M tokens blended (list price, 19 Nov 2026).
- Inkling 2 Small streams ~2× faster than Qwen4 Max (Artificial Analysis, 7-day median, 19 Nov 2026).
- Of the 16 benchmarks both models have, Qwen4 Max leads 13 (BenchLeader, 19 Nov 2026).
Which to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | Inkling 2 Small | 59 vs 55 category score; FrontierCode 2.0 (AA) 25.4% vs 26.4% | Medium |
| Reasoning | Qwen4 Max | 59 vs 52 category score; HLE-Diamond (AA) 44.9% vs 28.7% | Medium |
| Knowledge | Qwen4 Max | 59 vs 53 category score; AA-Omniscience v1.1 27.7 vs 16.1 | Medium |
| Chat quality (human preference) | – | Only Inkling 2 Small has an Arena rating (59 category score) | Not comparable |
| Lowest cost | Inkling 2 Small | $0.475 vs $2.25 per 1M blended; agentic coding session $0.022 vs $0.108 | – |
| Fastest first answer | Inkling 2 Small | 40.0 s (Inkling 2 Small) vs 139.2 s (Qwen4 Max) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Not enough shared benchmarks for a verdict.
- Inkling 2 Small is stronger in coding.
- Qwen4 Max is stronger in knowledge, reasoning.
- Inkling 2 Small is 4.7× cheaper ($0.475 vs $2.25 per 1M blended).
- Inkling 2 Small streams 2.0× faster (117 vs 59 tokens per second).
At a glance
| Spec | Inkling 2 Small | Qwen4 Max |
|---|---|---|
| Developer | Thinking Machines | Alibaba |
| Released | 5 Nov 2026 | 16 Nov 2026 |
| Context window (tokens) | 1,048,576 | 991,808 |
| Compared at | Inkling 2 Small | Qwen4 Max |
| Open weights | No | No |
| Price in / out / cached input per 1M | $0.300 / $1.00 / $0.030 | $1.50 / $4.50 / $0.071 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Inkling 2 Small | Qwen4 Max | Leader |
|---|---|---|---|
| Reasoning | 52.1 | 58.5 | Qwen4 Max |
| Coding | 59.4 | 55.1 | Inkling 2 Small |
| Agents & tools | 60.6 | – | not comparable |
| Knowledge | 53.1 | 59.0 | Qwen4 Max |
| Human preference | 58.7 | – | not comparable |
| Multimodal | 63.6 | – | not comparable |
| Long context | 66.4 | – | not comparable |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Not drawn, measured for fewer than two of these models: Agents & tools, Maths, Instruction following, Long context, Multimodal, Human preference.
Where sources disagree
Across the 16 benchmarks both have, Inkling 2 Small leads 3, Qwen4 Max leads 13. Every overall index that rates both puts Qwen4 Max ahead.
| Overall index | Inkling 2 Small | Qwen4 Max | Leader |
|---|---|---|---|
| AA Intelligence Index v5.0.1 | 30.6 | 42.9 | Qwen4 Max |
| Vals Index | 56.1 | 63.2† | Qwen4 Max |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | Inkling 2 Small | Qwen4 Max |
|---|---|---|
| Blended price per 1M tokens | $0.475 | $2.25 |
| Input / output per 1M | $0.300 / $1.00 | $1.50 / $4.50 |
| Cached input per 1M | $0.030 | $0.071 |
| Price per task: chat reply | $0.0004 | $0.0020 |
| Price per task: summarize a 30-page report | $0.0042 | $0.021 |
| Price per task: code edit | $0.0033 | $0.016 |
| Price per task: agentic coding session | $0.022 | $0.108 |
| Price per task: structured extraction | $0.0008 | $0.0039 |
| Agentic coding session, input mostly cached | $0.0099 | $0.044 |
| Cost per benchmark task (AA Index) | $1.18 | $3.27 |
| Output tokens per task (AA Index) | 150k | 105k |
| Of which reasoning | 88k | 69k |
| Time per task (AA Index) | 21 min | 30 min |
| Output speed | 117 tok/s | 59 tok/s |
| Time to first answer | 40.0 s | 139.2 s |
| Time to a complete answer, last measured 13 Nov 2026 | 163.6 s | – |
Per token is not per task: an average task of Artificial Analysis' Intelligence Index cost $1.18 on Inkling 2 Small and $3.27 on Qwen4 Max, and Inkling 2 Small wrote 1.4× as many output tokens for the same work, which a per-token price hides.
Each model is compared at its best-ranked reasoning-effort setting. Inkling 2 Small scores 57.6 and takes 40 s to its first answer token. Qwen4 Max and takes 139 s to its first answer token. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
89 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Inkling 2 Small | Qwen4 Max |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 57.6 | – |
| Agents & tools score | BenchLeader | 60.6 | – |
| Coding score | BenchLeader | 59.4 | 55.1 |
| Human preference score | BenchLeader | 58.7 | – |
| Knowledge score | BenchLeader | 53.1 | 59.0 |
| Long context score | BenchLeader | 66.4 | – |
| Multimodal score | BenchLeader | 63.6 | – |
| Reasoning score | BenchLeader | 52.1 | 58.5 |
| Blended price $/M | BenchLeader | $0.475 | $2.25 |
| Output speed | BenchLeader | 117 tok/s | 59 tok/s |
| Time to first answer | BenchLeader | 40.0 s | 139.2 s |
| Context window | BenchLeader | 1,048,576 tokens | 991,808 tokens |
| Reasoning | |||
| Arena Hard Prompts | Arena (LMArena) | 1439 | – |
| LiveBench 2.0 Reasoning | LiveBench | – | 92.2%† |
| Humanity's Last Exam (AA) | Artificial Analysis | 44.1% | – |
| HLE-Diamond (AA) | Artificial Analysis | 28.7% | 44.9% |
| ARC-AGI-1 | ARC Prize | 96.5%† | – |
| ARC-AGI-2 | ARC Prize | 81.7%† | – |
| ARC-AGI-3 | ARC Prize | 2.6%† | – |
| CritPt v2 hard subset (AA) | Artificial Analysis | 35.0% | 43.0% |
| MysteryMechanism | Vals AI | 25.2% | – |
| Arena Expert | Arena (LMArena) | 1491 | – |
| Coding | |||
| SciCode | SciCode | 52.3% | – |
| Arena Coding | Arena (LMArena) | 1486 | – |
| Arena WebDev | Arena (LMArena) | 1644 | 1750† |
| LiveBench 2.0 Coding | LiveBench | – | 69.8%† |
| ProgramBench | Vals AI | 0.5% | – |
| SWE-bench Multimodal | Vals AI | 43.1% | – |
| Vibe Code Bench v1.1 | Vals AI | 81.5% | 86.7%† |
| SWE Atlas v1.1 | Scale AI SEAL | 61.9% | – |
| SWE Atlas v1.1: Codebase QnA | Scale AI SEAL | 74.2% | – |
| SWE Atlas v1.1: Test Writing | Scale AI SEAL | 58.1% | – |
| SWE Atlas v1.1: Refactoring | Scale AI SEAL | 53.3% | – |
| SWE-Bench Pro V2 | Scale AI SEAL | 97.7% | – |
| SWE-Bench Pro V2 HARD | Scale AI SEAL | 86.3% | – |
| FrontierSWE | Epoch AI Benchmarking Hub | 25.1%† | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 13.1% | 48.5% |
Show all 89 benchmarks
| Metric | Publisher | Inkling 2 Small | Qwen4 Max |
|---|---|---|---|
| Coding | |||
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 81.3% | – |
| FrontierCode 2.0 (AA) | Artificial Analysis | 25.4% | 26.4% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 13.0%† | – |
| APEX-Agents | Mercor | 46.8% | – |
| Arena Agent | Arena (LMArena) | 1.5 | 7.3† |
| LiveBench Long Agentic | LiveBench | – | 46.6%† |
| LiveBench Browser | LiveBench | – | 80.7%† |
| MCP Atlas | Scale AI SEAL | 85.4% | – |
| GDPval-AA v2.1 | Artificial Analysis | 1630 | 1809 |
| τ³-Banking (AA) | Artificial Analysis | 40.4% | – |
| Analyst Agent (AA) | Artificial Analysis | 51.3% | 46.3% |
| APEX-Agents (AA) | Artificial Analysis | – | 59.1% |
| CyberBench | Vals AI | 72.7% | – |
| Finance Agent v2.1 | Vals AI | 67.5% | 66.7%† |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 21.7% | 21.3%† |
| Legal Research Bench | Vals AI | 46.6% | 41.4%† |
| SREBench | Vals AI | 1.9% | – |
| Tax Agent Bench | Vals AI | 38.9% | 40.4%† |
| Terminal-Bench 4.0 (Vals) | Vals AI | 12.6% | 48.5%† |
| Terminal-Bench Science | Vals AI | 2.9% | – |
| Arena Agent Task Completion | Arena (LMArena) | 2.4 | 8.8† |
| AutomationBench | Artificial Analysis | 45.6% | 72.6% |
| GDP.pdf (AA) | Artificial Analysis | 25.2% | 26.2% |
| Harvey LAB-AA v1.1 | Harvey | – | 9.4% |
| EnterpriseOps-Gym | Artificial Analysis | 46.8% | 48.5% |
| AA-Briefcase v1.1 | Artificial Analysis | 1527 | 1686 |
| Design Arena Agentic | Design Arena | 1306 | 1356 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 5.7% | 23.3% |
| Maths | |||
| ArXivMath | MathArena | 31.9% | – |
| BrokenArXiv | MathArena | 63.2% | – |
| Knowledge | |||
| LiveBench 2.0 Data Analysis | LiveBench | – | 83.4%† |
| LiveBench Research | LiveBench | – | 84.7%† |
| AA-Omniscience v1.1 | Artificial Analysis | 16.1 | 27.7 |
| BioMysteryBench | Vals AI | 70.0% | – |
| Excel Modeling Benchmark v1.1 | Vals AI | 69.5% | 70.8%† |
| MedCode | Vals AI | 44.1% | – |
| MedScribe | Vals AI | 90.2% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 43.6%† | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 33.8% | 43.7% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 73.3% | 71.6% |
| Instruction following | |||
| Arena Instruction Following | Arena (LMArena) | 1410 | – |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1412 | – |
| Arena Creative Writing | Arena (LMArena) | 1343 | – |
| Arena Multi-turn | Arena (LMArena) | 1412 | – |
| Arena Longer Queries | Arena (LMArena) | 1408 | – |
| Multimodal | |||
| LiveBench Design | LiveBench | – | 79.1%† |
| MMMU-Pro | Artificial Analysis | 81.6% | – |
| SAGE | Vals AI | 88.6% | – |
| Humanity's Sixth Sense | Scale AI SEAL | 33.3% | – |
| VIALS-AA | Handshake AI | 42.9% | – |
| Code2Video-AA | Artificial Analysis | 1020 | – |
| AA-3D | Artificial Analysis | 1027 | – |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 83.3% | – |
| MLCR | Artificial Analysis | 20.6% | 19.4% |
| Composite | |||
| LiveBench 2.0 | LiveBench | – | 73.1%† |
| AA Intelligence Index v5.0.1 | Artificial Analysis | 30.6 | 42.9 |
| AA Progress Index v1 | Artificial Analysis | 317.6 | – |
| Vals Index | Vals AI | 56.1 | 63.2† |
| Vals Finance Index | Vals AI | 61.5% | 62.0%† |
| Vals Science Index | Vals AI | 32.7% | – |
| Vals Legal Index | Vals AI | 73.9% | 70.7%† |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 16.2% | – |
| DistressBench | Scale AI SEAL | 82.3% | – |
| Metadata | |||
| AA Openness Index | Artificial Analysis | 38.9 | – |
Data as of 2026-11-19. Best configuration of each model; every score links to its source on the model pages.
What changed
- Inkling 2 Small: 2 score revisions.
- Inkling 2 Small: 3 new results (AA Progress Index v1, MedCode, MedScribe).
- Inkling 2 Small: 1 score revision.
- Inkling 2 Small: 1 score revision; Inkling 2 Small cached input price moved from $0.060/M to $0.030/M (50% down).
- Inkling 2 Small: 1 score revision.
- Inkling 2 Small: BenchLeader Index 58.4 → 57.9.
- Inkling 2 Small: 2 new results (FORTRESS, Design Arena Agentic).
- Inkling 2 Small: 2 new results (MysteryMechanism, Vals Science Index); 1 score revision.
- Inkling 2 Small: 3 new results (SciCode, APEX-Agents, MLCR).
- Inkling 2 Small: 8 new results (Arena Text, Arena Hard Prompts, Arena Coding…); 2 score revisions.
Related comparisons
- Inkling 2 Small vs Gemini 4 Carbon
- Inkling 2 Small vs Claude Opus 5.5
- Inkling 2 Small vs GPT-6 Astra
- Inkling 2 Small vs Claude Fable 5.1
- Qwen4 Max vs Qwen3.8 Max
- Inkling 2 Small vs Qwen3.7 Plus
- Inkling 2 Small vs GPT-6.1 Sol
- Inkling 2 Small vs Claude Fable 5
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Inkling 2 Small vs Qwen4 Max: questions
- Inkling 2 Small vs Qwen4 Max: which is cheaper, and will that change?
- Inkling 2 Small is cheaper today: $0.475 against $2.25 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Inkling 2 Small or Qwen4 Max better for coding?
- Inkling 2 Small scores higher in coding (59 vs 55 on the category index, where 50 is average). On FrontierCode 2.0 (AA), the top shared coding benchmark, it is 25.4% for Inkling 2 Small against 26.4% for Qwen4 Max.
- Which is faster, Inkling 2 Small or Qwen4 Max?
- Inkling 2 Small streams ~2× faster (117 vs 59 output tokens per second, Artificial Analysis 7-day median).
- Which is better for long documents?
- Inkling 2 Small accepts more context: 1,048,576 against 991,808 tokens for Qwen4 Max.
- Which has lower latency?
- Inkling 2 Small answers first: 40.0 s against 139.2 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
More questions (2)
- Do independent benchmarks agree on Inkling 2 Small vs Qwen4 Max?
- Across the 16 benchmarks both have, Inkling 2 Small leads 3, Qwen4 Max leads 13. Every overall index that rates both puts Qwen4 Max ahead.
- When was Qwen4 Max released?
- Qwen4 Max was released on 16 Nov 2026.