Claude Sonnet 4.6 vs Inkling 2 Small
Full profiles: Claude Sonnet 4.6 (Anthropic) and Inkling 2 Small (Thinking Machines): every benchmark, price and rank.
Updated · Data from 10 sources
As of 14 Nov 2026, Claude Sonnet 4.6 and Inkling 2 Small are neck and neck on quality (58.3 vs 57.9 on the BenchLeader Index, inside the ±4.7 margin), and Inkling 2 Small is 13× cheaper at $0.475 vs $6.00 per 1M tokens blended. Choose Claude Sonnet 4.6 for reasoning (64 vs 53), knowledge (60 vs 54), and human preference (61 vs 59) and fast first answers (0.7 s vs 72.4 s); choose Inkling 2 Small for agents & tools (61 vs 51), coding (60 vs 55), and multimodal (64 vs 62), where it also streams ~2× faster.
Key facts (as of 14 Nov 2026)
- Claude Sonnet 4.6 scores 58.3 on the BenchLeader Index and Inkling 2 Small scores 57.9 (BenchLeader, 14 Nov 2026).
- Claude Sonnet 4.6 costs $6.00 per 1M tokens blended (list price, 14 Nov 2026).
- Inkling 2 Small costs $0.475 per 1M tokens blended (list price, 14 Nov 2026).
- Inkling 2 Small streams ~2× faster than Claude Sonnet 4.6 (Artificial Analysis, 7-day median, 14 Nov 2026).
- Of the 14 benchmarks both models have, Inkling 2 Small leads 7 (BenchLeader, 14 Nov 2026).
- On Arena WebDev, Claude Sonnet 4.6 is rated 1522 and Inkling 2 Small 1647 (LMArena, as published by 14 Nov 2026).
Claude Sonnet 4.6 benchmarks Anthropic
Inkling 2 Small benchmarks Thinking MachinesAdd a third modelWhich 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 | 60 vs 55 category score; Arena WebDev 1647 vs 1522 | Medium |
| Agents & tool use | Inkling 2 Small | 61 vs 51 category score; MCP Atlas 85.4% vs 69.5% | High |
| Reasoning | Claude Sonnet 4.6 | 64 vs 53 category score; Arena Hard Prompts 1503 vs 1438 | High |
| Knowledge | Claude Sonnet 4.6 | 60 vs 54 category score | Medium |
| Images and other media | Inkling 2 Small | 64 vs 62 category score | Medium |
| Chat quality (human preference) | Claude Sonnet 4.6 | 61 vs 59 on Arena votes | Medium |
| Lowest cost | Inkling 2 Small | $0.475 vs $6.00 per 1M blended; agentic coding session $0.022 vs $0.240 | – |
| Fastest first answer | Claude Sonnet 4.6 | 0.7 s (Claude Sonnet 4.6) vs 72.4 s (Inkling 2 Small) 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
- Claude Sonnet 4.6 and Inkling 2 Small are neck and neck on quality (58.3 vs 57.9, within the ±4.7 margin).
- Claude Sonnet 4.6 is stronger in human preference, knowledge, reasoning.
- Inkling 2 Small is stronger in agents & tools, coding, multimodal.
- Inkling 2 Small is 13× cheaper ($0.475 vs $6.00 per 1M blended).
- Inkling 2 Small streams 1.7× faster (76 vs 44 tokens per second).
At a glance
| Spec | Claude Sonnet 4.6 | Inkling 2 Small |
|---|---|---|
| Developer | Anthropic | Thinking Machines |
| Released | 17 Feb 2026 | 5 Nov 2026 |
| Context window (tokens) | 1,000,000 | 1,048,576 |
| Compared at | Claude Sonnet 4.6 | Inkling 2 Small |
| Open weights | No | No |
| Price in / out / cached input per 1M | $3.00 / $15.00 / $0.300 | $0.300 / $1.00 / – |
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 | Claude Sonnet 4.6 | Inkling 2 Small | Leader |
|---|---|---|---|
| Reasoning | 63.9 | 52.8 | Claude Sonnet 4.6 |
| Coding | 55.2 | 60.3 | Inkling 2 Small |
| Agents & tools | 51.3 | 61.1 | Inkling 2 Small |
| Maths | 63.1 | – | not comparable |
| Knowledge | 59.8 | 53.5 | Claude Sonnet 4.6 |
| Human preference | 60.5 | 58.6 | Claude Sonnet 4.6 |
| Multimodal | 62.4 | 64.0 | Inkling 2 Small |
| 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: Maths, Instruction following, Long context.
Model, click to add or removeIndex
Where sources disagree
Across the 14 benchmarks both have, Claude Sonnet 4.6 leads 6, Inkling 2 Small leads 7 and 1 is a tie. Every overall index that rates both puts Claude Sonnet 4.6 ahead.
| Overall index | Claude Sonnet 4.6 | Inkling 2 Small | Leader |
|---|---|---|---|
| BenchLeader Index | 58.3 | 57.9 | neck and neck |
| Arena Text | 1472 | 1412 | Claude Sonnet 4.6 |
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 | Claude Sonnet 4.6 | Inkling 2 Small |
|---|---|---|
| Blended price per 1M tokens | $6.00 | $0.475 |
| Input / output per 1M | $3.00 / $15.00 | $0.300 / $1.00 |
| Cached input per 1M | $0.300 | – |
| Price per task: chat reply | $0.0057 | $0.0004 |
| Price per task: summarize a 30-page report | $0.045 | $0.0042 |
| Price per task: code edit | $0.041 | $0.0033 |
| Price per task: agentic coding session | $0.240 | $0.022 |
| Price per task: structured extraction | $0.0090 | $0.0008 |
| Agentic coding session, input mostly cached | $0.118 | – |
| Cost per benchmark task (AA Index) | – | $1.18 |
| Output tokens per task (AA Index) | – | 150k |
| Of which reasoning | – | 88k |
| Time per task (AA Index) | – | 33 min |
| Output speed | 44 tok/s | 76 tok/s |
| Time to first answer | 0.7 s | 72.4 s |
| Time to a complete answer | – | 181.0 s |
Each model is compared at its best-ranked reasoning-effort setting. Claude Sonnet 4.6 scores 58.3 and takes 1 s to its first answer token; Claude Sonnet 4.6 (max) scores 54.0 in 123 s, Claude Sonnet 4.6 (high) scores 50.3 in 1 s. Inkling 2 Small scores 57.9 and takes 72 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
132 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 | Claude Sonnet 4.6 | Inkling 2 Small |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 58.3 | 57.9 |
| Agents & tools score | BenchLeader | 51.3 | 61.1 |
| Coding score | BenchLeader | 55.2 | 60.3 |
| Human preference score | BenchLeader | 60.5 | 58.6 |
| Knowledge score | BenchLeader | 59.8 | 53.5 |
| Maths score | BenchLeader | 63.1 | – |
| Multimodal score | BenchLeader | 62.4 | 64.0 |
| Reasoning score | BenchLeader | 63.9 | 52.8 |
| Long context score | BenchLeader | – | 66.4 |
| Blended price $/M | BenchLeader | $6.00 | $0.475 |
| Output speed | BenchLeader | 44 tok/s | 76 tok/s |
| Time to first answer | BenchLeader | 0.7 s | 72.4 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,048,576 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 87.4% | – |
| Arena Hard Prompts | Arena (LMArena) | 1503 | 1438 |
| LiveBench 1.0 Reasoning | LiveBench | 84.8%† | – |
| GPQA Diamond (AA) | Artificial Analysis | 87.5%† | – |
| Humanity's Last Exam (AA) | Artificial Analysis | 33.6%† | 44.1% |
| HLE-Diamond (AA) | Artificial Analysis | – | 28.7% |
| GPQA Diamond (Vals) | Vals AI | 85.6% | – |
| ARC-AGI-1 | ARC Prize | 86.0%† | 96.5%† |
| ARC-AGI-2 | ARC Prize | 58.3%† | 81.7%† |
| CritPt | Artificial Analysis | 3.1%† | – |
| CritPt v2 hard subset (AA) | Artificial Analysis | – | 35.0% |
| MysteryMechanism | Vals AI | – | 25.2% |
| Chess Puzzles | Epoch AI Benchmarking Hub | 23.0%† | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 14.0% | – |
| LMCA | Epoch AI Benchmarking Hub | 46.5%† | – |
| DTBench | Epoch AI Benchmarking Hub | 89.9%† | – |
| ForecastBench | ForecastBench | 64.0% | – |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 75.2% | – |
| SciCode | SciCode | 46.8%† | 52.3% |
| WeirdML | WeirdML | 66.1%† | – |
| FrontierCode | Cognition | 24.3% | – |
| Arena Coding | Arena (LMArena) | 1528 | 1485 |
| Arena WebDev | Arena (LMArena) | 1522 | 1647 |
| LiveBench 1.0 Coding | LiveBench | 79.3%† | – |
| SciCode (AA) | Artificial Analysis | 50.1%† | 52.3% |
Show all 132 benchmarks
| Metric | Publisher | Claude Sonnet 4.6 | Inkling 2 Small |
|---|---|---|---|
| Coding | |||
| LiveCodeBench | Vals AI | 82.1% | – |
| SWE-bench (Vals) | Vals AI | 77.4% | – |
| Code Migration | Vals AI | 39.9% | – |
| ProgramBench | Vals AI | 0.5% | 0.5% |
| SWE-bench Multimodal | Vals AI | – | 43.1% |
| Vibe Code Bench v1.1 | Vals AI | 51.5% | 81.5% |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 31.2% | – |
| SWE Atlas: Refactoring | Scale AI SEAL | 32.2% | – |
| SWE Atlas: Test Writing | Scale AI SEAL | 31.8% | – |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 29.9%† | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 1327.3† | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 3.0%† | 13.1% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 71.2%† | 81.3% |
| Design Arena Website | Design Arena | 1277 | 1335 |
| FrontierCode 2.0 (AA) | Artificial Analysis | – | 25.4% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 53.4% | 13.0%† |
| OSWorld-Verified 2.0 | OSWorld | 8.3%† | – |
| APEX-Agents | Mercor | 43.0%† | 46.8% |
| Arena Agent | Arena (LMArena) | – | 2.5 |
| LiveBench Agentic Coding | LiveBench | 42.6%† | – |
| Terminal-Bench Hard | Artificial Analysis | 53.0%† | – |
| τ²-Bench Telecom (AA) | Artificial Analysis | 75.7%† | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | 57.3% | – |
| MCP Atlas | Scale AI SEAL | 69.5% | 85.4% |
| GDPval-AA v2.1 | Artificial Analysis | – | 1630 |
| τ³-Banking (AA) | Artificial Analysis | 34.4%† | 40.4% |
| ITBench SRE (AA) | Artificial Analysis | 39.8%† | – |
| Analyst Agent (AA) | Artificial Analysis | 20.0%† | 51.3% |
| APEX-Agents (AA) | Artificial Analysis | 28.0%† | – |
| CyberBench | Vals AI | – | 72.7% |
| Finance Agent v2 | Vals AI | 51.0% | – |
| Finance Agent v2.1 | Vals AI | – | 67.5% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | – | 21.7% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 5.0% | – |
| Legal Research Bench | Vals AI | 38.5% | 46.6% |
| SkillsBench | Vals AI | 49.0% | – |
| SREBench | Vals AI | – | 1.9% |
| Tax Agent Bench | Vals AI | 25.8% | 38.9% |
| Terminal-Bench 2.0 (Vals) | Vals AI | 59.5% | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | – | 12.6% |
| Terminal-Bench Science | Vals AI | – | 2.9% |
| Arena Search | Arena (LMArena) | 1201 | – |
| DeepResearch Bench | DeepResearch Bench | 54.9%† | – |
| ExploitBench | Epoch AI Benchmarking Hub | 23.6% | – |
| GBAEval | Epoch AI Benchmarking Hub | 48.8% | – |
| Vending-Bench 2 | Andon Labs | 7204.1 | – |
| AutomationBench | Artificial Analysis | 20.1%† | 45.6% |
| GDP.pdf (AA) | Artificial Analysis | 15.8%† | 25.2% |
| EnterpriseOps-Gym | Artificial Analysis | – | 46.8% |
| AA-Briefcase v1.1 | Artificial Analysis | 1062† | 1527 |
| Design Arena Agentic | Design Arena | 1181 | 1310 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 5.7% |
| Maths | |||
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 85.8% | – |
| ProofBench | Vals AI | 45.0%† | – |
| LiveBench Mathematics | LiveBench | 87.0%† | – |
| AIME (Vals) | Vals AI | 92.3% | – |
| Arena Maths | Arena (LMArena) | 1461 | – |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 32.8%† | – |
| LiveBench 1.0 Data Analysis | LiveBench | 78.0%† | – |
| AA-Omniscience v1.1 | Artificial Analysis | – | 16.1 |
| MMLU-Pro | Vals AI | 87.3% | – |
| LegalBench | Vals AI | 82.1% | – |
| CorpFin | Vals AI | 65.3% | – |
| TaxEval | Vals AI | 77.1% | – |
| MedQA | Vals AI | 92.1% | – |
| BioMysteryBench | Vals AI | – | 70.0% |
| CaseLaw v2 | Vals AI | 64.0% | – |
| Excel Modeling Benchmark | Vals AI | 60.1% | – |
| Excel Modeling Benchmark v1.1 | Vals AI | – | 69.5% |
| Public Benefits Bench v1.1 | Vals AI | 62.5% | – |
| DrugDiscoveryBench | Scale AI SEAL | 31.3%† | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 18.0%† | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | – | 33.8% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | – | 73.3% |
| Global-MMLU-Lite (AA) | Artificial Analysis | 90.5%† | – |
| Instruction following | |||
| LiveBench Language | LiveBench | 76.1%† | – |
| LiveBench Instruction Following | LiveBench | 63.2%† | – |
| IFBench | Artificial Analysis | 56.6%† | – |
| Arena Instruction Following | Arena (LMArena) | 1475 | 1410 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1472 | 1412 |
| EQ-Bench 4 | EQ-Bench | 1197 | – |
| Arena Creative Writing | Arena (LMArena) | 1449 | 1345 |
| Arena Multi-turn | Arena (LMArena) | 1480 | 1415 |
| Arena Longer Queries | Arena (LMArena) | 1492 | 1410 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1283 | – |
| MMMU-Pro | Artificial Analysis | 73.3%† | 81.6% |
| MMMU-Pro (Vals) | Vals AI | 83.6% | – |
| MortgageTax | Vals AI | 67.7% | – |
| SAGE | Vals AI | – | 88.6% |
| GeoBench | GeoBench | 73.0% | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 6.7% | – |
| Code2Video-AA | Artificial Analysis | – | 1020 |
| AA-3D | Artificial Analysis | – | 1027 |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 80.0%† | 83.3% |
| Arena Document | Arena (LMArena) | 1478 | – |
| MLCR | Artificial Analysis | 7.8%† | 20.6% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 152.9 | – |
| LiveBench 1.0 | LiveBench | 73.0%† | – |
| AA Intelligence Index v5.0 | Artificial Analysis | – | 30.6 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 30.1† | – |
| AA Progress Index v1 | Artificial Analysis | 293.6† | – |
| Vals Index | Vals AI | – | 56.1 |
| Vals Finance Index | Vals AI | – | 61.5% |
| Vals Science Index | Vals AI | – | 32.7% |
| Vals Multimodal Index | Vals AI | 60.6% | – |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | – | 16.2% |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 38.9 |
Data as of 2026-11-14. Best configuration of each model; every score links to its source on the model pages.
What changed
- 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.
- Claude Sonnet 4.6: 1 new result (GeoBench).
- 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.
- Inkling 2 Small: BenchLeader Index 57.5 → 58.4.
- Claude Sonnet 4.6: 1 new result (Arena Search).
- Inkling 2 Small: 1 new result (Arena WebDev).
Related comparisons
- Claude Sonnet 4.6 vs Gemini 4 Carbon
- Claude Sonnet 4.6 vs GPT-6.1 Sol
- 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
- Claude Sonnet 4.6 vs Claude Sonnet 5.5
- Claude Sonnet 4.6 vs Muse Spark 1.3
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Claude Sonnet 4.6 vs Inkling 2 Small: questions
- Is Claude Sonnet 4.6 better than Inkling 2 Small?
- Claude Sonnet 4.6 and Inkling 2 Small are neck and neck on quality (58.3 vs 57.9, within the ±4.7 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 14 Nov 2026, so check the category scores for your use.
- Claude Sonnet 4.6 vs Inkling 2 Small: which is cheaper, and will that change?
- Inkling 2 Small is cheaper today: $0.475 against $6.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Sonnet 4.6 or Inkling 2 Small better for coding?
- Inkling 2 Small scores higher in coding (60 vs 55 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1522 for Claude Sonnet 4.6 against 1647 for Inkling 2 Small.
- Which is faster, Claude Sonnet 4.6 or Inkling 2 Small?
- Inkling 2 Small streams ~2× faster (76 vs 44 output tokens per second, Artificial Analysis 7-day median).
- Is Claude Sonnet 4.6 or Inkling 2 Small better for agents and tool use?
- Inkling 2 Small scores higher in agentic tasks (61 vs 51 on the category index, where 50 is average). On MCP Atlas, the top shared agents & tools benchmark, it is 69.5% for Claude Sonnet 4.6 against 85.4% for Inkling 2 Small.
More questions (4)
- Which is better for long documents?
- Both take about 1M tokens: Claude Sonnet 4.6 1,000,000, Inkling 2 Small 1,048,576.
- Which has lower latency?
- Claude Sonnet 4.6 answers first: 0.7 s against 72.4 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.
- Do independent benchmarks agree on Claude Sonnet 4.6 vs Inkling 2 Small?
- Across the 14 benchmarks both have, Claude Sonnet 4.6 leads 6, Inkling 2 Small leads 7 and 1 is a tie. Every overall index that rates both puts Claude Sonnet 4.6 ahead.
- When was Inkling 2 Small released?
- Inkling 2 Small was released on 5 Nov 2026.