Skip to content

Inkling

Thinking Machines LabOpen weightsHybridVerifiedReleased Jul 15, 2026Input: text, image, audio
Compare with Inkling-SmallSource scores ↗OpenRouter ↗
Overall score
52.6
Rank 90 of 222 · 42.3–62.9
Context window
1M
1,000,000 tokens
Price per 1M tokens
$1.87 in · $4.68 out
Source: BenchLM
Blended price
$2.57
3:1 input:output tokens
Output speed
—
Not measured yet

Category scores

Scores per capability area with the model's rank among ranked models in that category.

  1. 1Instruction following#3687.6
  2. 2Reasoning#1278.1
  3. 3Knowledge#6568.1
  4. 4Multimodal#3567.8
  5. 5Agentic & tool use#7762.7
  6. 6Math#1158.3
  7. 7Coding#8454.3
Higher is better. #N is the model's rank in that category, where it is eligible to be ranked.

Benchmark results (19)

As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.

Agentic & tool use

Terminal-Bench 2.163.8#49 of 68
BrowseComp77.1#26 of 50
MCP Atlas74.1#20 of 42
CWE-bench v137.0#11 of 11
Design Arena Agentic Web Dev1257#1 of 1

Multimodal

Design Arena Website1224#53 of 91
MMMU-Pro73.5#38 of 48
CharXiv82.0#20 of 42
CharXiv w/o tools78.1#12 of 14

Coding

SWE-bench Verified77.6#30 of 98
SWE-bench Pro54.3#50 of 69
FrontierSWE v24.1#20 of 20

Knowledge

GPQA87.9#35 of 117
HLE46.0#23 of 69
HLE w/o tools30.0#32 of 43

Math

FrontierMath v2 (Tiers 1-3)33.333#52 of 64
FrontierMath v2 (Tier 4)4.878#46 of 51
AIME2697.1#3 of 29

Instruction following

IFBench79.8#10 of 42

Pricing by source

USD per 1M tokens.

SourceInputOutputCached in
List price (BenchLM)$1.87$4.68$0.37
OpenRouter thinkingmachines/inkling$1.00$4.05$0.17

Thinking Machines Lab's Tinker pricing docs list the 64K Inkling tier at $1.87 prefill / $0.374 cached prefill / $4.68 sample per million tokens during a limited-time 50% discount. The hosted 256K tier costs $3.74 / $0.748 cached / $9.36, while the released checkpoint supports up to 1M tokens when self-hosted. BenchLM stores the lower-context hosted tier as the headline API rate and the model's architectural context ceiling separately.

Arena ratings

Human-preference ratings from LMArena. Best-rated variant shown.

Text1441#74 · 40.6K votes
WebDev (coding)1414#90 · 15.8K votes

Open weights

From the Hugging Face repository.

Parameters
952B
Licence
apache-2.0
Downloads (30 days)
1.3M

Speed

Measured by us through OpenRouter (median of trials).

Not measured yet.

Related models

Same family

Similar score