Inkling
Thinking Machines Lab · released
Thinking Machines' first model: a very large open-weight model that reads text, images and audio. It is built for tool-using agents, calibrated answers and fine-tuning on the lab's Tinker platform; the lab itself says it is not the strongest model available.
Best for: Fine-tuning and self-hosting · Audio and image understanding · Calibrated answers
- Intelligence Index
- 25.0
- #30 of 38 · at xHigh effort
- Price per 1M tokens
- $1 in $4.05 out
- Blended $1.76 at 3:1
- Context window
- 1M
- Output speed
- 151 tokens/s
- Median, measured by Artificial Analysis
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall31 · #31 of 38
Based on 3 of 3 benchmarks
- Coding12 · #34 of 39
Based on 5 of 5 benchmarks
- Agents & knowledge work23 · #31 of 40
Based on 6 of 6 benchmarks
- Reasoning & science30 · #30 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability61 · #25 of 39
Based on 2 of 2 benchmarks
- Vision54 · #24 of 31
Based on 1 of 2 benchmarks
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#30 of 38· xHigh effort· Artificial Analysis
- Vals Index28.7%
#25 of 26· $1.46 per test· Vals AI
- Arena · Text1441 ±5
#25 of 33· 35,204 votes· Arena
Coding
#27 of 38· xHigh effort· Artificial Analysis
- SciCode47.0%
#29 of 38· xHigh effort· Artificial Analysis
- Arena · WebDev1412 ±6
#30 of 34· 14,166 votes· Arena
- Vibe Code Bench19.2%
#25 of 26· Vals AI
- Code Migration11.8%
#26 of 26· Vals AI
Agents & knowledge work
- GDPval-AA28.9
#27 of 39· xHigh effort· Artificial Analysis
- Arena · Agent-10.9 ±1
#28 of 29· Arena
- Finance Agent v246.6%
#24 of 26· Vals AI
#25 of 26· Vals AI
- Legal Research Bench28.4%
#24 of 26· Vals AI
- Tax Agent Bench43.5%
#24 of 26· Vals AI
Reasoning & science
- Humanity's Last Exam31.9%
#30 of 39· xHigh effort· Artificial Analysis
- CritPt5.4%
#27 of 39· xHigh effort· Artificial Analysis
- ARC-AGI-2saturating36.5%
#16 of 16· $0.64 per task· ARC Prize
- ARC-AGI-3Not tested
- GPQA Diamondsaturating87.2%
#18 of 27· xHigh effort· Artificial Analysis
Knowledge & reliability
#19 of 39· xHigh effort· Artificial Analysis
- AA-LCR77.3%
#29 of 39· xHigh effort· Artificial Analysis
Vision
- MMMU-Pro73.5%
#21 of 27· xHigh effort· Artificial Analysis
- Arena · VisionNot tested
Reported by Thinking Machines Lab
Numbers the lab published at launch, run on its own harness and settings. Useful context, but not independently verified and never used in the rankings on this site.
- GPQA Diamond · effort 0.9987.2%
- MMMU-Pro · effort 0.9973.5%
- ARC-AGI-2 · effort 0.9936.5%
Specifications
- Released
- Jul 15, 2026
- Availability
- Generally available
- Context window
- 1,000,000 tokens
- Max output
- —
- Accepts
- Text, Images, Audio
- Knowledge cutoff
- Not published
- Weights
- Open
- Licence
- Apache 2.0
- Parameters
- 975B total · 41B active
- Reasoning control
- Adjustable thinking effort
- API model ID
thinkingmachines/Inkling
Pricing
- Input / 1M tokens
- $1
- Cached input / 1M
- $0.17
- Output / 1M tokens
- $4.05
- Blended (3:1)
- $1.76
Serverless price on the lab's Tinker platform, which currently serves a 256K context.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.