Inkling (batch)
Thinking Machines · released Jul 17, 2026
Inkling is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab, with 41B active parameters out of 975B total. It is designed for general-purpose reasoning, coding, agentic and tool-use systems,...
Specification
- Context window
- 524K
- Max output
- —
- Knowledge cutoff
- Not stated
- Parameters
- 952.4B
- Licence
- apache-2.0
- Serving providers
- 1
- Moderated
- No
Intelligence
42.3
91th percentile
Coding
52.1
Coding Index
Agentic
34.1
Agentic Index
Output speed
81 t/s
Median across providers
Latency
2.21s
Time to first token
Cost per task
$0.19
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- GPQA Diamond87.2%
- AA-LCR (long context)73.3%
- SciCode46.1%
- Humanity's Last Exam31.9%
- τ²-bench (Telecom)29.1%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| GPQA Diamond | 87.2% |
| AA-LCR (long context) | 73.3% |
| SciCode | 46.1% |
| Humanity's Last Exam | 31.9% |
| τ²-bench (Telecom) | 29.1% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- GPT-5.2OpenAI43.3
- Kimi K2.7 CodeMoonshot AI43.0
- MiMo-V2.5-ProXiaomi42.9
- Inkling (batch)Thinking Machines42.3
- InklingThinking Machines42.3
- Hy3Tencent42.2
- Nex-N2-ProNex AGI41.7
- Inkling SmallThinking Machines41.2
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| GPT-5.2 | 43.3 |
| Kimi K2.7 Code | 43.0 |
| MiMo-V2.5-Pro | 42.9 |
| Inkling (batch) | 42.3 |
| Inkling | 42.3 |
| Hy3 | 42.2 |
| Nex-N2-Pro | 41.7 |
| Inkling Small | 41.2 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $0.5
- Output / 1M tokens
- $2.02
- Cached input / 1M
- $0.085
- Blended 3:1
- $0.881
One task, estimated
$0.19
- Input tokens
- 50,000
- Output tokens
- 80,000
- Profile
- Reasoning
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Serving providers
Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.
Arena
Head-to-head generation quality
Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.
- Overall Elo
- 1,232
- Win rate
- 44.6%
- Strongest at
- Websites
- Tournaments
- 1,204
Elo by category
Dot position on a 1,150–1,250 scale · Elo has no meaningful zero
- 1,232
- 1,203
- 1,200
- 1,171
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| Websites | 1232 | 44.6% |
| Game development | 1203 | 40.5% |
| Data visualisation | 1200 | 43.4% |
| 3D scenes | 1171 | 37.7% |