Skip to content
aiqbench

Thinky · model passport

Thinking Machines: Inkling

inkling
  • Open weights
  • Apache 2.0 licence
  • 952.4B params
  • 524,288 context
  • Released 2026-07-14(HF repo created)
12/27axes covered
Overall · LMArena
1,441#87 of 396
Best percentile
79.5Coding
Cohort standing
58.2median 59.8

What the data says

Benchmarked on 12 of 27 tracked axes. Every figure below is the model's own published result on that board.

Weakest axes

  • 5.5 pct — Agent Arena (-0.144)

Uneven: 74 percentile points between its best and worst axis — 79.5 pct on Coding versus 5.5 pct on Agent Arena.

  • Coding79.5 pct1,493Verify →
  • Hard prompts79 pct1,462Verify →
  • Overall78.3 pct1,441Verify →
  • Instruction following76.8 pct1,425Verify →
  • GPQA Diamond76.5 pct88.3%Verify →
  • AIME (math)70.5 pct88.9%Verify →
  • Factuality Arena66.3 pct1,451Verify →
  • ARC-AGI-256.2 pct36.5%Verify →
  • SimpleBench50.6 pct50.0%Verify →
  • WebDev Arena43.2 pct1,414Verify →
  • SimpleQA Verified41.5 pct40.3%Verify →
  • Agent Arena5.5 pct-0.144Verify →
  • Vision ArenaNot benchmarked
  • Search ArenaNot benchmarked
  • Document ArenaNot benchmarked
  • FrontierMathNot benchmarked
  • Humanity's Last ExamNot benchmarked
  • LiveBenchNot benchmarked
  • BFCL v4 (function calling)Not benchmarked
  • SWE-bench Verified (live board)Not benchmarked
  • Terminal-Bench 4.0Not benchmarked
  • MMLU-Pro (Artificial Analysis)Not benchmarked
  • LiveCodeBench (Artificial Analysis)Not benchmarked
  • IFBench (Artificial Analysis)Not benchmarked
  • HHEM Hallucination (Vectara)Not benchmarked
  • SWE-bench VerifiedNot benchmarked
  • Terminal-BenchNot benchmarked

Percentiles are the model's standing within each board's own population. Axes marked "not benchmarked" have no published result for this model — that is an absence of data, not a low score.

How to run it

Run it yourself

Repositorythinkingmachines/Inkling
Parameters952.4B
ArchitectureInklingForConditionalGeneration
LicenceApache 2.0

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "thinkingmachines/Inkling"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

Ollama

We don't sync Ollama tags, so we won't print a command we can't verify — a wrong one would pull the wrong model. Search the Ollama library →

Unsloth fine-tuning recipes are model-specific and we don't derive them here. Unsloth docs →

Use it hosted

API model IDthinkingmachines/inkling
Context length524,288 tokens
Max output471,859 tokens
Acceptstext, image, audio
Input price$1 per 1M tokens
Cached input price$0.17 per 1M tokens
Output price$4.05 per 1M tokens

API parameters this endpoint supports

Frequency penaltyInclude reasoningLogit biasMax tokensMin-pPresence penaltyReasoningReasoning effortRepetition penaltySeedStop sequencesTemperatureTool choiceTool callingTop-kTop-p

Example request

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"thinkingmachines/inkling","messages":[{"role":"user","content":"Hello"}]}'

View on OpenRouter →

Identifiers, limits and prices are synced from the upstream source shown above. Any field we haven't synced is left out entirely rather than guessed.

  • inkling stands at 58.2 (raw mean 60.3) — measured on 12 of 27 axes.

    standingtemplateVerify →

  • inkling stands at 58.2 versus a cohort median of 59.8 across 196 models released within 180 days — below the peer median.

    cohorttemplateVerify →

Every claim is a computed fact from the current board data, phrased by a low-temperature model call and rejected by a validation gate if any number is not in the source row. How the leaderboards work →