Inkling-Small
Thinking Machines Lab · released
A smaller sibling of Inkling with the same text, image and audio inputs. It is cheaper and faster and edges out Inkling on reasoning and coding benchmarks, but trails it on factual recall.
Best for: Low-cost multimodal inference · Fine-tuning · Speech and audio understanding
- Intelligence Index
- 25.7
- #28 of 38
- Price per 1M tokens
- $0.30 in $1.2 out
- Blended $0.52 at 3:1
- Context window
- 1M
- Output speed
- 205 tokens/s
- $0.09 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall25 · #33 of 38
Based on 3 of 3 benchmarks
- Coding13 · #32 of 39
Based on 5 of 5 benchmarks
- Agents & knowledge work17 · #33 of 40
Based on 6 of 6 benchmarks
- Reasoning & science36 · #26 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability53 · #30 of 39
Based on 2 of 2 benchmarks
- Vision39 · #27 of 31
Based on 2 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
#28 of 38· Artificial Analysis
- Vals Index25.5%
#26 of 26· $0.28 per test· Vals AI
- Arena · Text1405 ±5
#31 of 33· 27,070 votes· Arena
Coding
#27 of 38· Artificial Analysis
- SciCode49.7%
#25 of 38· Artificial Analysis
- Arena · WebDev1409 ±9
#31 of 34· 5,372 votes· Arena
- Vibe Code Bench19.1%
#26 of 26· Vals AI
- Code Migration13.3%
#25 of 26· Vals AI
Agents & knowledge work
- GDPval-AA31.2
#26 of 39· Artificial Analysis
- Arena · Agent-10.3 ±1.4
#27 of 29· Arena
- Finance Agent v241.3%
#26 of 26· Vals AI
#26 of 26· Vals AI
- Legal Research Bench25.5%
#25 of 26· Vals AI
- Tax Agent Bench41.5%
#25 of 26· Vals AI
Reasoning & science
- Humanity's Last Exam33.3%
#29 of 39· Artificial Analysis
- CritPt8.3%
#26 of 39· Artificial Analysis
- ARC-AGI-2saturating40.1%
#15 of 16· xHigh effort· $0.23 per task· ARC Prize
- ARC-AGI-3Not tested
- GPQA Diamondsaturating89.5%
#17 of 27· Artificial Analysis
Knowledge & reliability
#28 of 39· Artificial Analysis
- AA-LCR75.7%
#32 of 39· Artificial Analysis
Vision
- MMMU-Pro74.0%
#20 of 27· Artificial Analysis
- Arena · Vision1206 ±9
#18 of 19· 6,904 votes· Arena
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 0.9989.5%
- MMMU-Pro · effort 0.9974%
- ARC-AGI-2 · effort 0.9940.1%
Specifications
- Released
- Jul 30, 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
- 276B total · 12B active
- Reasoning control
- Effort minimal to xhigh
- API model ID
thinkingmachines/Inkling-Small
Pricing
- Input / 1M tokens
- $0.30
- Cached input / 1M
- $0.06
- Output / 1M tokens
- $1.2
- Blended (3:1)
- $0.52
- Measured cost / task
- $0.09
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.