Skip to content
ModelCap

Inkling 2 Mini

Official publisher releaseNewPreliminaryOpen weights15.6h #63 → #64

Thinking Machines·#64 of 374 on the current board·cheaper than #63 Grok 4.7

Inkling 2 Mini is a language model from Thinking Machines. It ranks #64 of 374 on ModelCap's current board, with a preliminary Index score of 63.7, partly measured on public benchmarks and partly modeled. Its score range of 52.4–75.1 spans positions #27–#111. It is listed at $0.075 per million input tokens and $0.30 per million output tokens across 2 providers. It reads up to 1M tokens of context, and its weights are openly downloadable.

Figures as of 21 Nov 2026, 19:03 UTC · download the public dataset · methodology

ModelCap Index · Current board
63.7#64
Partly measured on public benchmarks and partly modeled · range 52.4–75.1
Evidence as of 20 Nov 2026, 20:43 UTC
Market Gravity
27.8
Secondary market signal
Input
$0.075
per 1M tokens
Output
$0.30
per 1M tokens
Context
1M
tokens
Providers
2
endpoints

Providers

Uptime measured over the last 30 minutes
ProviderInputOutputContextQuantUptime
PrimeIntellectfp4Cheapest$0.075$0.301Mfp4100.0%
Io Netnvfp4$0.08$0.35262Knvfp4100.0%

Hugging Face Inference Providers

Exact repository identity

Provider economics and telemetry published by the Hugging Face Router. These are operational signals, not benchmark results.

ProviderInputOutputContextTTFTThroughputFeatures
primeintellect
live
$0.075$0.301M661 ms103.0 tok/stools · structured
io-net
live
$0.08$0.35262K544 ms119.0 tok/stools · structured

ModelCap Index · Current board

Methodology
63.7
Partly measured on public benchmarks and partly modeled
Current board
#64 of 374

Preliminary. Measured on 1 public benchmark so far, with limited support. Expect the score and position to move as more boards report. What this label means

Score range
52.4–75.1
Rank range
#27–#111
Evidence as of
20 Nov 2026, 20:43 UTC
Benchmarks used
ARC-AGI-2 Verified
Technical details
Public rank basis
Measured · blended (measured)
Evidence label
Measured · blended · publisher-corpus-prior over 217 held-out anchors (14%); 1 specialist-board observation at 2.1% support (10%); launch card against 2 resolved peers on 12 rows (76%); exceeds every named peer on 6 of 12 rows; optimism haircut 0 from cross-lab-probe; shrunk 1.4 toward the measured corpus
Index support
2.1% · weak
Identity binding
Exact catalogue product · thinkingmachines/thinkingmachines/inkling-2-mini · aggregates disclosed benchmark configurations · endpoint configuration not claimed · artifact metadata thinkingmachines/Inkling-2-Mini@ed40e7da625f594fc9c4990f5f66be24873c22d7 (not the evaluation revision)
Index observations used
1
Qualified Index evidence point
63.7
How this position was produced · fused launch placement

No general-family board lists this model yet. Its point estimate is the precision-weighted mean of every channel that spoke for it; each channel’s share is the weight it carried. Independent measurements replace this placement as boards list the model.

Fused estimate
63.7 ± 8.9
Independent share
10%
Shrink toward corpus
-1.4
Peer ceiling share
50%
Optimism haircut
None required · cross lab probe, 1 probe
Method
precision-weighted-v1
Channels fused into this model’s launch placement
ChannelEstimateShare
Publisher corpus prior · not independent
publisher-corpus-prior over 217 held-out anchors
54.8 ± 24.014%
Specialist board measurement
1 specialist-board observation at 2.1% support
39.8 ± 27.610%
Launch card against resolved peers · not independent
launch card against 2 resolved peers on 12 rows
68.6 ± 10.276%
  • · Launch card row automationbench-aa-v-1-0-6: inside the resolved peer range
  • · Launch card row deepswe-v-1-1: claims above every resolved peer
  • · Launch card row frontiercode-v-2-0: claims above every resolved peer
  • · Launch card row hle-diamond: inside the resolved peer range
  • · Launch card row mcp-atlas: claims above every resolved peer
  • · Launch card row mmmu-pro: inside the resolved peer range
  • · Launch card row non-hallucination-rate: inside the resolved peer range
  • · Launch card row swebench-multilingual: claims above every resolved peer
  • · Launch card row swebench-pro-v-2: claims above every resolved peer
  • · Launch card row swebench-verified: claims above every resolved peer
  • · Launch card row terminal-bench-v-2-1: inside the resolved peer range
  • · Launch card row terminal-bench-v-4-0: inside the resolved peer range
Bayesian Evidence Score · measured evidence input only

BES normalizes admitted public benchmark evidence for measured Index rows. Its legacy score and observables prior do not define the public language rank.

Observed capability
39.8
Evidence status
provisional · 2% mass
General preference
—
Coding
—
Agents & tools
—
Reasoning
39.8
Evidence breadth
3%
Evidence coverage
2%
Market and catalogue signals · secondary, never the language Index rank
Usage (OpenRouter popularity)
14.5
Liquidity (providers × uptime)
50.0
Open reach (HF downloads)
15.4
Surface (context / tools / modalities)
100.0
Freshness
99.7
  • · OpenRouter popularity #147
  • · 2 live providers
  • · HF 30d downloads 7944
  • · Catalogue: context 1048576, reasoning, tools, modalities Text/Image
  • · First seen 2026-11-21

These adoption and deployment observations remain context only. They do not change this model's ModelCap Index score or rank.

BES evidence selected · measured input only

The bounded 0–100 capability score blends each source’s competitive placement with its published achievement, then combines capability families. Evidence breadth adds a modest uncertainty adjustment; price and popularity are not part of benchmark-led rank.

Deployment readiness

Metadata index
83.1
strong · 94% metadata coverage
Not capability
Artifact reproducibility
85.0
Access & legal clarity
100.0
Deployability
100.0
Evaluation provenance
37.9

Not yet published: the base model it derives from, pinned evaluation dataset versions.

A metadata completeness and deployability index—not a safety certification, quality grade, or production-readiness claim.

Market activity

Venue coverage
Market Gravity
27.8
OpenRouter weekly popularity
#147 of 342

Market Gravity uses OpenRouter’s full-catalogue popularity order. Sparse Vercel top-ten observations appear only when published and do not affect the score.

Arena preference

No matched Arena rating

This model is visible because it is available in the live catalogue, but it has no trustworthy community result yet. ModelCap does not infer a quality score from price, features, or market activity.

Market Gravity breakdown

Usage55%
14.5

OpenRouter weekly popularity across the full model catalogue.

Liquidity25%
50.0

Independent providers versus the model's open or closed cohort, weighted by uptime.

Open reach15%
15.4

Hugging Face 30-day downloads for open models; neutral for closed models.

Freshness5%
99.7

Time since first public availability, on a six-month half-life.

Specification

Model ID
thinkingmachines/inkling-2-mini
Context window
1M tokens
Declared output cap
66K tokens

OpenRouter catalogue or provider-endpoint declaration. Limits can vary by provider and prompt length; a model-wide maximum has not been independently verified. OpenRouter metadata

Inputs
Text, Image
Outputs
Text
Tokenizer
Other
Cached input tokens
$0.015 / 1M
First seen on OpenRouter
21 Nov 2026

Capabilities

  • Supported: Reasoning
  • Supported: Tool use
  • Supported: Structured output
  • Supported: Response format
  • Not supported: Moderated

Weights & access

Open weights. Downloadable weights are published under an OSI or free-culture license in the current repository metadata. This describes weight availability, not whether the full training stack qualifies as Open Source AI.

Access
Open weights
Licence
apache-2.0
Pinned revision
ed40e7da625f
Parameters
26.4B
Architecture
Inkling2LoopForConditionalGeneration
Model type
inkling2_loop
Hugging Face downloads (30d)
8K
Downloads (all time)
8K
Likes
2K
Repository updated
20 Nov 2026, 22:55 UTC

Compare Inkling 2 Mini with

Pick any model

Benchmark leaderboards featuring Inkling 2 Mini

All leaderboards
Public benchmark boards on which Inkling 2 Mini has a published result
LeaderboardScoreSource rankConfigurationPublished
ARC-AGI-2 leaderboardARC Prize Foundation49.3%#132 of 303X-High20 Nov 2026
Artificial Analysis Intelligence IndexArtificial Analysis20.1#83 of 126Reasoning21 Nov 2026

Alternatives to Inkling 2 Mini, by the numbers

Ranked neighbours

Models within 3 places of #64 on the ModelCap Index.

Cheaper at a similar or higher Index

Index score within 3 points of, or above, this model's, with a lower listed blended price (3:1 input:output).

Other ranked Thinking Machines models

The same publisher's other current models with a public Index position.

Inkling 2 Mini: common questions

How does Inkling 2 Mini rank among AI models?

Inkling 2 Mini holds ModelCap Index position #64 of 374 ranked language models as of 21 November 2026, with an Index score of 63.7 (interval 52.4–75.1); evidence: measured · blended. The Index combines public benchmark boards with published uncertainty; the methodology page explains the weighting.

How much does Inkling 2 Mini cost per 1M tokens?

Inkling 2 Mini is listed at $0.075 per 1M input tokens and $0.30 per 1M output tokens as of 21 November 2026 across 2 providers. Prices come from the OpenRouter catalogue and re-observe on every refresh. For 1,000 requests using 2,000 input and 500 output tokens each, the listed rates imply $0.30 total (2M input + 0.5M output). This excludes caching, batch discounts, prompt-length tiers, tool charges and retries; verify the chosen endpoint's rate and limits before budgeting.

What is the context window of Inkling 2 Mini?

Inkling 2 Mini has a published context window of 1M tokens (1,048,576). The snapshot lists a 65,536 token catalogue-declared route output cap. OpenRouter catalogue or provider-endpoint declaration. Limits can vary by provider and prompt length; a model-wide maximum has not been independently verified. The providers table lists each endpoint's context limit.

Is Inkling 2 Mini open-weight?

Open weights (apache-2.0). Downloadable weights are published under an OSI or free-culture license in the current repository metadata. This describes weight availability, not whether the full training stack qualifies as Open Source AI. The repository is thinkingmachines/Inkling-2-Mini on Hugging Face.

Which API providers serve Inkling 2 Mini?

2 providers serve Inkling 2 Mini through OpenRouter as of 21 November 2026: PrimeIntellect and Io Net. Each provider's price, context limit, quantization and measured uptime are in the providers table above.

Which benchmarks has Inkling 2 Mini been evaluated on?

Inkling 2 Mini has published results on 2 public boards tracked by ModelCap as of 21 November 2026: ARC-AGI-2 49.3% (#132 of 303) and AA Intelligence Index 20.1 (#83 of 126). Each board page ranks every tracked model on that benchmark. The ModelCap Index scores the results its methodology admits, with published uncertainty, and shows reference boards without scoring them.

How does Inkling 2 Mini compare with Inkling?

Inkling 2 Mini ranks #64 (Index 63.7) and Inkling ranks #65 (Index 63.6) on the ModelCap Index as of 21 November 2026. The head-to-head page lines up their benchmarks, listed prices, context windows and provider counts side by side.

When was Inkling 2 Mini released?

Inkling 2 Mini first appeared in the catalogue on 21 Nov 2026. It is the current version in its family. Rank and price movements since then are recorded on the site's changes feed.