Compare / head-to-head
InklingvsInkling-Small
Inkling and Inkling-Small split their 21 shared benchmarks evenly. Inkling-Small is 3.4x cheaper per token. Both offer a 1M-token context window.
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
10 – 10
even split
Cheaper per token
Inkling-Small
3.4x cheaper, input + output
Larger context
Tie
both 1M tokens
Providers
Thinking Machines Lab
same provider
Inkling
Thinking Machines Lab · released 2026-07-15
textvisionaudio
Inkling-Small
Thinking Machines Lab · released 2026-07-30
textvisionaudio
Quality
Benchmark matrix
21 shared · 17 only Inkling · 7 only Inkling-Small| Benchmark | Inkling | Inkling-Small | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 21 | ||||
| SWE-bench Verified | 77.6% ↗ | 80.2% ↗ | -2.6 pt | Inkling-Small |
| GPQA Diamond | 88.3% ↗epoch run | 88.5% ↗epoch run | -0.2 pt | Inkling-Small |
| AIME 2026 | 97.1% ↗ | 95.5% ↗ | +1.6 pt | Inkling |
| LMArena Elo | 1441.4 ↗ | 1405.2 ↗ | +36.2 | Inkling |
| CharXiv RQ | 78.1% ↗ | 77.4% ↗ | +0.7 pt | Inkling |
| CharXiv RQ (with python) | 82% ↗ | 81.3% ↗ | +0.7 pt | Inkling |
| Chess Puzzles (Epoch AI run) | 21% ↗epoch run | 18% ↗epoch run | +3 pt | Inkling |
| FrontierMath Tier 4 v2 (Epoch AI run) | 4.9% ↗epoch run | 17.1% ↗epoch run | -12.2 pt | Inkling-Small |
| FrontierMath Tiers 1-3 v2 (Epoch AI run) | 33.3% ↗epoch run | 46.3% ↗epoch run | -13 pt | Inkling-Small |
| IFBench | 79.8% ↗ | 82.2% ↗ | -2.4 pt | Inkling-Small |
| LMArena Agent (LMArena) | -0.1086 ↗ | -0.1033 ↗ | 0 | Tie |
| LMArena WebDev (LMArena) | 1412.6 ↗ | 1408.2 ↗ | +4.4 | Inkling |
| MMAU | 77.2% ↗ | 77% ↗ | +0.2 pt | Inkling |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | 88.9% ↗epoch run | 90% ↗epoch run | -1.1 pt | Inkling-Small |
| SAGE (Vals AI) | 36.6% ↗ | 32.7% ↗ | +3.9 pt | Inkling |
| SimpleQA Verified | 43.9% ↗ | 20.6% ↗ | +23.3 pt | Inkling |
| SWE-bench Pro (public) | 54.3% ↗ | 55.9% ↗ | -1.6 pt | Inkling-Small |
| Terminal-Bench 2.1 (best harness) | 63.8% ↗ | 64.7% ↗ | -0.9 pt | Inkling-Small |
| Terminal-Bench 4.0 (Vals AI) | 0.5% ↗ | 0% ↗ | +0.5 pt | Inkling |
| Toolathlon Verified | 45.5% ↗ | 54.4% ↗ | -8.9 pt | Inkling-Small |
| Toolathlon-Verified (HKUST) | 45.5% ↗ | 54.4% ↗ | -8.9 pt | Inkling-Small |
| Only Inkling reports · 17 | ||||
| APEX-Agents (Mercor) | 33.8% ↗ | not reported | — | — |
| Audio MC | 56.6% ↗ | not reported | — | — |
| BrowseComp (w/ ctx management) | 77.1% ↗ | not reported | — | — |
| FrontierSWE V2 (Proximal Labs) | 4.1% ↗ | not reported | — | — |
| Global-MMLU-Lite | 88.7% ↗ | not reported | — | — |
| LiveBench Agentic Coding (LiveBench) | 49.4% ↗ | not reported | — | — |
| LiveBench Coding (LiveBench) | 71% ↗ | not reported | — | — |
| LiveBench Data Analysis (LiveBench) | 72.8% ↗ | not reported | — | — |
| LiveBench Instruction Following (LiveBench) | 70.1% ↗ | not reported | — | — |
| LiveBench Language (LiveBench) | 73.5% ↗ | not reported | — | — |
| LiveBench Mathematics (LiveBench) | 88.4% ↗ | not reported | — | — |
| LiveBench Reasoning (LiveBench) | 78.3% ↗ | not reported | — | — |
| MCP Atlas | 76% ↗ | not reported | — | — |
| SimpleBench (SimpleBench) | 50% ↗ | not reported | — | — |
| tau2-bench Banking Knowledge (Sierra) | 25% ↗ | not reported | — | — |
| VoiceBench | 91.4% ↗ | not reported | — | — |
| WeirdML (Håvard Tveit Ihle) | 32.3% ↗ | not reported | — | — |
| Only Inkling-Small reports · 7 | ||||
| ARC-AGI-1 | not reported | 84% ↗ | — | — |
| ARC-AGI-2 | not reported | 40.1% ↗ | — | — |
| BrowseComp (with context management) | not reported | 77.4% ↗ | — | — |
| HMMT Feb 2026 | not reported | 90.2% ↗ | — | — |
| LMArena Vision (LMArena) | not reported | 1206.3 ↗ | — | — |
| MCP Atlas (all) | not reported | 79.2% ↗ | — | — |
| Mystery Game Puzzles (Epoch AI run) | not reported | 6% ↗epoch run | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Inkling minus Inkling-Small in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Inkling | Inkling-Small | Edge |
|---|---|---|---|
| Context window | 1M tokens | 1M tokens | Tie |
| Max output | — | — | — |
| Input price / 1M | $1 ↗ | $0.3 ↗ | Inkling-Small |
| Output price / 1M | $4.05 ↗ | $1.2 ↗ | Inkling-Small |
| Cached input / 1M | $0.17 ↗ | $0.06 ↗ | Inkling-Small |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | $5.05 | $1.5 | Inkling-Small |
| Modalities | text · vision · audio | text · vision · audio | Tie |
| Released | 2026-07-15 | 2026-07-30 | — |
| Cited benchmark scores | 39 | 29 | — |
Reliability
Thinking Machines Lab status
Live from /statusMore matchups
Inkling vs …
Models sharing the most benchmarksMore matchups
Inkling-Small vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Inkling and Inkling-Small on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.