GPT-5.6 Luna vs Inkling 2 Small
Full profiles: GPT-5.6 Luna (OpenAI) and Inkling 2 Small (Thinking Machines): every benchmark, price and rank.
Updated · Data from 12 sources
As of 16 Nov 2026, GPT-5.6 Luna and Inkling 2 Small are neck and neck on quality (58.2 vs 57.9 on the BenchLeader Index, inside the ±5.2 margin), and they cost about the same at $0.450 vs $0.475 per 1M tokens blended. Choose GPT-5.6 Luna for human preference (64 vs 59) and fast first answers (28.1 s vs 49.0 s); choose Inkling 2 Small for multimodal (64 vs 60), agents & tools (61 vs 57), and coding (60 vs 56).
Key facts (as of 16 Nov 2026)
- GPT-5.6 Luna scores 58.2 on the BenchLeader Index and Inkling 2 Small scores 57.9 (BenchLeader, 16 Nov 2026).
- GPT-5.6 Luna costs $0.450 per 1M tokens blended (list price, 16 Nov 2026).
- Inkling 2 Small costs $0.475 per 1M tokens blended (list price, 16 Nov 2026).
- Of the 20 benchmarks both models have, Inkling 2 Small leads 13 (BenchLeader, 16 Nov 2026).
- On Arena WebDev, GPT-5.6 Luna is rated 1519 and Inkling 2 Small 1647 (LMArena, as published by 16 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 | 60 vs 56 category score; Arena WebDev 1647 vs 1519 | Medium |
| Agents & tool use | Inkling 2 Small | 61 vs 57 category score; τ³-Banking (AA) 40.4% vs 28.7% | Medium |
| Reasoning | Inkling 2 Small | 53 vs 52 category score; Arena Hard Prompts 1438 vs 1475 | Medium |
| Long documents | Either | 66 vs 66 category score; AA-LCR v1.1 81.7% vs 83.3% | Either |
| Images and other media | Inkling 2 Small | 64 vs 60 category score; MMMU-Pro 81.6% vs 78.5% | Medium |
| Chat quality (human preference) | GPT-5.6 Luna | 64 vs 59 on Arena votes | Medium |
| Lowest cost | Either | $0.450 vs $0.475 per 1M blended; agentic coding session $0.017 vs $0.022 | – |
| Fastest first answer | GPT-5.6 Luna | 28.1 s (GPT-5.6 Luna (xhigh)) vs 49.0 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
- GPT-5.6 Luna (xhigh) and Inkling 2 Small are neck and neck on quality (58.2 vs 57.9, within the ±5.2 margin).
- GPT-5.6 Luna (xhigh) is stronger in human preference.
- Inkling 2 Small is stronger in agents & tools, coding, long context, multimodal, reasoning.
- They cost about the same ($0.450 per 1M blended).
At a glance
| Spec | GPT-5.6 Luna | Inkling 2 Small |
|---|---|---|
| Developer | OpenAI | Thinking Machines |
| Released | 9 Jul 2026 | 5 Nov 2026 |
| Context window (tokens) | 1,050,000 | 1,048,576 |
| Compared at | GPT-5.6 Luna (xhigh) | Inkling 2 Small |
| Open weights | No | No |
| Price in / out / cached input per 1M | $0.200 / $1.20 / $0.020 | $0.300 / $1.00 / $0.030 |
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 | GPT-5.6 Luna (xhigh) | Inkling 2 Small | Leader |
|---|---|---|---|
| Reasoning | 51.8 | 52.8 | Inkling 2 Small |
| Coding | 56.4 | 60.3 | Inkling 2 Small |
| Agents & tools | 57.1 | 61.1 | Inkling 2 Small |
| Maths | 65.7 | – | not comparable |
| Knowledge | – | 53.5 | not comparable |
| Human preference | 63.7 | 58.6 | GPT-5.6 Luna |
| Multimodal | 59.9 | 64.0 | Inkling 2 Small |
| Long context | 65.5 | 66.4 | level |
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, Knowledge, Instruction following.
Where sources disagree
Across the 20 benchmarks both have, GPT-5.6 Luna leads 7, Inkling 2 Small leads 13. Independent sources split: GPT-5.6 Luna leads on Arena Text; Inkling 2 Small leads on AA Intelligence Index v5.0.1.
| Overall index | GPT-5.6 Luna (xhigh) | Inkling 2 Small | Leader |
|---|---|---|---|
| BenchLeader Index | 58.2 | 57.9 | neck and neck |
| AA Intelligence Index v5.0.1 | 23.8† | 30.6 | Inkling 2 Small |
| Arena Text | 1453 | 1412 | GPT-5.6 Luna |
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 | GPT-5.6 Luna (xhigh) | Inkling 2 Small |
|---|---|---|
| Blended price per 1M tokens | $0.450 | $0.475 |
| Input / output per 1M | $0.200 / $1.20 | $0.300 / $1.00 |
| Cached input per 1M | $0.020 | $0.030 |
| Price per task: chat reply | $0.0004 | $0.0004 |
| Price per task: summarize a 30-page report | $0.0031 | $0.0042 |
| Price per task: code edit | $0.0030 | $0.0033 |
| Price per task: agentic coding session | $0.017 | $0.022 |
| Price per task: structured extraction | $0.0006 | $0.0008 |
| Agentic coding session, input mostly cached | $0.0087 | $0.0099 |
| Cost per benchmark task (AA Index) | – | $1.18 |
| Output tokens per task (AA Index) | – | 150k |
| Of which reasoning | – | 88k |
| Time per task (AA Index) | – | 25 min |
| Output speed | 110 tok/s | 100 tok/s |
| Time to first answer | 28.1 s | 49.0 s |
| Time to a complete answer, last measured 13 Nov 2026 | 101.7 s | 163.6 s |
Each model is compared at its best-ranked reasoning-effort setting. GPT-5.6 Luna (xhigh) scores 58.2 and takes 28 s to its first answer token; GPT-5.6 Luna (max) scores 56.2 in 59 s, GPT-5.6 Luna (high) scores 54.7 in 11 s. Inkling 2 Small scores 57.9 and takes 49 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
149 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 | GPT-5.6 Luna (xhigh) | Inkling 2 Small |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 58.2 | 57.9 |
| Agents & tools score | BenchLeader | 57.1 | 61.1 |
| Coding score | BenchLeader | 56.4 | 60.3 |
| Human preference score | BenchLeader | 63.7 | 58.6 |
| Long context score | BenchLeader | 65.5 | 66.4 |
| Maths score | BenchLeader | 65.7 | – |
| Multimodal score | BenchLeader | 59.9 | 64.0 |
| Reasoning score | BenchLeader | 51.8 | 52.8 |
| Knowledge score | BenchLeader | – | 53.5 |
| Blended price $/M | BenchLeader | $0.450 | $0.475 |
| Output speed | BenchLeader | 110 tok/s | 100 tok/s |
| Time to first answer | BenchLeader | 28.1 s | 49.0 s |
| Context window | BenchLeader | 1,050,000 tokens | 1,048,576 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 91.6%† | – |
| SimpleBench | SimpleBench | 46.8% | – |
| Arena Hard Prompts | Arena (LMArena) | 1475 | 1438 |
| LiveBench 1.0 Reasoning | LiveBench | 85.6%† | – |
| GPQA Diamond (AA) | Artificial Analysis | 89.5% | – |
| Humanity's Last Exam (AA) | Artificial Analysis | 37.0% | 44.1% |
| HLE-Diamond (AA) | Artificial Analysis | 20.9%† | 28.7% |
| GPQA Diamond (Vals) | Vals AI | 91.7%† | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 49.1%† | – |
| ARC-AGI-1 | ARC Prize | 87.7% | 96.5%† |
| ARC-AGI-2 | ARC Prize | 47.6% | 81.7%† |
| ARC-AGI-3 | ARC Prize | 0.0% | – |
| CritPt | Artificial Analysis | 20.6% | – |
| CritPt v2 hard subset (AA) | Artificial Analysis | 30.5%† | 35.0% |
| MysteryMechanism | Vals AI | 14.4%† | 25.2% |
| Arena Expert | Arena (LMArena) | 1490 | 1488 |
| Chess Puzzles | Epoch AI Benchmarking Hub | 43.0%† | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 21.0%† | – |
| LMCA | Epoch AI Benchmarking Hub | 48.5% | – |
| DTBench | Epoch AI Benchmarking Hub | 89.1% | – |
| Coding | |||
| SciCode | SciCode | 50.5% | 52.3% |
| WeirdML | WeirdML | 60.9%† | – |
| FrontierCode | Cognition | 39.8%† | – |
| Arena Coding | Arena (LMArena) | 1503 | 1485 |
| Arena WebDev | Arena (LMArena) | 1519 | 1647 |
Show all 149 benchmarks
| Metric | Publisher | GPT-5.6 Luna (xhigh) | Inkling 2 Small |
|---|---|---|---|
| Coding | |||
| LiveBench 1.0 Coding | LiveBench | 82.9%† | – |
| IOI | Vals AI | 61.8%† | – |
| SWE-bench (Vals) | Vals AI | 93.0%† | – |
| Code Migration | Vals AI | 44.5%† | – |
| ProgramBench | Vals AI | 0.0%† | 0.5% |
| SWE-bench Multimodal | Vals AI | – | 43.1% |
| Vibe Code Bench 1-100 | Vals AI | 22.6%† | – |
| Vibe Code Bench v1.1 | Vals AI | 77.1%† | 81.5% |
| 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% |
| MirrorCode | Epoch AI Benchmarking Hub | 21.1%† | – |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 61.9%† | – |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 56.9% | – |
| CursorBench | Cursor | 33.0% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 1667.4† | – |
| WeirdML v3 | WeirdML | 5.2% | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 3.5% | 13.1% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 77.9% | 81.3% |
| Startrise Frontend | Startrise | 65.0%† | – |
| GameBench 2 Overall | DuelLab GameBench 2 | 40.2† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 54 | – |
| Design Arena Website | Design Arena | 1248† | 1335 |
| FrontierCode 2.0 (AA) | Artificial Analysis | 17.6%† | 25.4% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 17.6%† | 13.0%† |
| APEX-Agents | Mercor | 43.0%† | 46.8% |
| Arena Agent | Arena (LMArena) | – | 2.5 |
| LiveBench Agentic Coding | LiveBench | 48.4%† | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | 79.0%† | – |
| MCP Atlas | Scale AI SEAL | – | 85.4% |
| GDPval-AA v2.1 | Artificial Analysis | 1407 | 1630 |
| τ³-Banking (AA) | Artificial Analysis | 28.7% | 40.4% |
| ITBench SRE (AA) | Artificial Analysis | 40.3%† | – |
| Analyst Agent (AA) | Artificial Analysis | – | 51.3% |
| APEX-Agents (AA) | Artificial Analysis | 35.8%† | – |
| CyberBench | Vals AI | 73.6%† | 72.7% |
| Finance Agent v2 | Vals AI | 55.0%† | – |
| 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 | 1.3%† | – |
| Legal Research Bench | Vals AI | 36.5%† | 46.6% |
| SkillsBench | Vals AI | 60.5%† | – |
| SREBench | Vals AI | 1.5%† | 1.9% |
| Tax Agent Bench | Vals AI | 24.8%† | 38.9% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 11.6%† | 12.6% |
| Terminal-Bench Science | Vals AI | 0.0%† | 2.9% |
| Arena Agent Task Completion | Arena (LMArena) | – | 3.3 |
| BALROG | BALROG | 45.6%† | – |
| Vending-Bench 2 | Andon Labs | 4094.7† | – |
| AutomationBench | Artificial Analysis | 42.6% | 45.6% |
| GDP.pdf (AA) | Artificial Analysis | 23.8% | 25.2% |
| EnterpriseOps-Gym | Artificial Analysis | 40.8%† | 46.8% |
| AA-Briefcase v1.1 | Artificial Analysis | 1267 | 1527 |
| Design Arena Agentic | Design Arena | 1149† | 1315 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 3.3%† | 5.7% |
| FACTS Search | 87.0%† | – | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 82.1%† | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 61.0%† | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 98.3%† | – |
| ProofBench | Vals AI | 60.0%† | – |
| LiveBench Mathematics | LiveBench | 87.2%† | – |
| ArXivMath | MathArena | – | 31.9% |
| BrokenArXiv | MathArena | – | 63.2% |
| Arena Maths | Arena (LMArena) | 1476 | – |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 41.0%† | – |
| LiveBench 1.0 Data Analysis | LiveBench | 78.0%† | – |
| AA-Omniscience v1.1 | Artificial Analysis | -17.1† | 16.1 |
| MMLU-Pro | Vals AI | 86.0%† | – |
| LegalBench | Vals AI | 84.0%† | – |
| CorpFin | Vals AI | 64.2%† | – |
| TaxEval | Vals AI | 76.2%† | – |
| BioMysteryBench | Vals AI | 61.5%† | 70.0% |
| Excel Modeling Benchmark | Vals AI | 67.1%† | – |
| Excel Modeling Benchmark v1.1 | Vals AI | – | 69.5% |
| MedCode | Vals AI | 42.4%† | – |
| MedScribe | Vals AI | 84.4%† | – |
| Public Benefits Bench v1.1 | Vals AI | 61.2%† | – |
| Access to Justice Benchmark | Vals AI | 65.6%† | – |
| Financial Coaching Bench | Vals AI | 69.3%† | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 22.7%† | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 39.1%† | 33.8% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 7.7%† | 73.3% |
| FACTS Parametric | 50.6%† | – | |
| Instruction following | |||
| LiveBench Language | LiveBench | 72.6%† | – |
| LiveBench Instruction Following | LiveBench | 60.1%† | – |
| Arena Instruction Following | Arena (LMArena) | 1443 | 1410 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1453 | 1412 |
| EQ-Bench 4 | EQ-Bench | 1146† | – |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1832† | – |
| Arena Creative Writing | Arena (LMArena) | 1411 | 1345 |
| Arena Multi-turn | Arena (LMArena) | 1458 | 1415 |
| Arena Longer Queries | Arena (LMArena) | 1452 | 1410 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1256 | – |
| MMMU-Pro | Artificial Analysis | 78.5% | 81.6% |
| MMMU-Pro (Vals) | Vals AI | 85.0%† | – |
| MortgageTax | Vals AI | 67.3%† | – |
| SAGE | Vals AI | – | 88.6% |
| Humanity's Sixth Sense | Scale AI SEAL | 21.6%† | 33.3% |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 22.6%† | – |
| Furniture Assembly | Epoch AI Benchmarking Hub | 42.5%† | – |
| VIALS-AA | Handshake AI | – | 42.9% |
| Code2Video-AA | Artificial Analysis | – | 1020 |
| AA-3D | Artificial Analysis | – | 1027 |
| FACTS Multimodal | 37.4%† | – | |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 81.7% | 83.3% |
| Arena Document | Arena (LMArena) | 1460 | – |
| MLCR | Artificial Analysis | 15.0% | 20.6% |
| FACTS Grounding | 52.5%† | – | |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 156.7† | – |
| LiveBench 1.0 | LiveBench | 73.6%† | – |
| AA Intelligence Index v5.0.1 | Artificial Analysis | 23.8† | 30.6 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 34.6 | – |
| AA Progress Index v1 | Artificial Analysis | 305.8 | – |
| Vals Index | Vals AI | – | 56.1 |
| Vals Finance Index | Vals AI | – | 61.5% |
| Vals Science Index | Vals AI | 25.3%† | 32.7% |
| Vals Multimodal Index | Vals AI | 69.1%† | – |
| Vals Legal Index | Vals AI | – | 73.9% |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | – | 16.2% |
| DistressBench | Scale AI SEAL | 72.7%† | 82.3% |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 38.9 |
Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- Inkling 2 Small: 1 score revision; Inkling 2 Small cached input price moved from $0.060/M to $0.030/M (50% down).
- GPT-5.6 Luna (xhigh): 2 new results (GDPval-AA v2.1, AA-Briefcase v1.1).
- 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.
- GPT-5.6 Luna (xhigh): 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.
- Inkling 2 Small: BenchLeader Index 57.5 → 58.4.
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
- GPT-5.6 Luna vs GPT-6.1 Sol
- GPT-5.6 Luna vs Gemini 4 Argon
- GPT-5.6 Luna vs Claude Sonnet 5.5
- GPT-5.6 Luna vs GPT-5.5
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
GPT-5.6 Luna vs Inkling 2 Small: questions
- Is GPT-5.6 Luna better than Inkling 2 Small?
- GPT-5.6 Luna (xhigh) and Inkling 2 Small are neck and neck on quality (58.2 vs 57.9, within the ±5.2 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
- GPT-5.6 Luna vs Inkling 2 Small: which is cheaper, and will that change?
- GPT-5.6 Luna is cheaper today: $0.450 against $0.475 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is GPT-5.6 Luna or Inkling 2 Small better for coding?
- Inkling 2 Small scores higher in coding (60 vs 56 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1519 for GPT-5.6 Luna against 1647 for Inkling 2 Small.
- Which is faster, GPT-5.6 Luna or Inkling 2 Small?
- GPT-5.6 Luna streams faster (110 vs 100 output tokens per second, Artificial Analysis 7-day median).
- Is GPT-5.6 Luna or Inkling 2 Small better for agents and tool use?
- Inkling 2 Small scores higher in agentic tasks (61 vs 57 on the category index, where 50 is average). On τ³-Banking (AA), the top shared agents & tools benchmark, it is 28.7% for GPT-5.6 Luna against 40.4% for Inkling 2 Small.
More questions (4)
- Which is better for long documents?
- They are level on long-context benchmarks (66 vs 66). Both take about 1.1M tokens: GPT-5.6 Luna 1,050,000, Inkling 2 Small 1,048,576.
- Which has lower latency?
- GPT-5.6 Luna answers first: 28.1 s against 49.0 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 GPT-5.6 Luna vs Inkling 2 Small?
- Across the 20 benchmarks both have, GPT-5.6 Luna leads 7, Inkling 2 Small leads 13. Independent sources split: GPT-5.6 Luna leads on Arena Text; Inkling 2 Small leads on AA Intelligence Index v5.0.1.
- When was Inkling 2 Small released?
- Inkling 2 Small was released on 5 Nov 2026.