BenchLeader

Inkling-Small

Best configuration ranks #184 of 610 on the BenchLeader Index at 55.0 ±4.2. Last measured 8 Sept 2026. Released 15 Jul 2026.

Blended price
$0.638/M
$0.450 in · $1.20 out
Output speed
150 tok/s
First answer
15 s
first token 0.33 s
Context
1M
How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.
  1. Overall index55
  2. Reasoning60
  3. Coding56
  4. Agents & tools49
  5. Maths34
  6. Knowledge60
  7. Human preference59
  8. Multimodal58
  9. Long context64
  10. Composite62

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.

EffortIndexRankSpeedFirst answerChat reply costCategories
no reasoning150 tok/s15 s$0.0006Reasoning 31
minimal150 tok/s15 s$0.0006Reasoning 39
low150 tok/s15 s$0.0006Reasoning 40
medium150 tok/s15 s$0.0006Reasoning 45
high150 tok/s15 s$0.0006Reasoning 52
xhigh48.5#322150 tok/s15 s$0.0006Knowledge 32 · Maths 52 · Reasoning 59
defaultbest55.0#184150 tok/s15 s$0.0005Agents & tools 49 · Coding 56 · Composite 62 · Human preference 59 · Knowledge 60 · Long context 64 · Maths 34 · Multimodal 58 · Reasoning 60

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmarkno reasoningminimallowmediumhighxhighdefaultSourceTrend
GPQA Diamond88.5%#48Epoch AI Benchmarking Hub
LMArena Hard Prompts1430#120LMArena
GPQA Diamond (AA)not in index89.5%#74Artificial Analysis
Humanity's Last Exam (AA)not in index33.3%#92Artificial Analysis
GPQA Diamond (Vals)not in index83.6%#55Vals AI
ARC-AGI-15.5%#17448.5%#12452.5%#12167.0%#9978.0%#8584.0%#77ARC Prize
ARC-AGI-20.0%#1744.0%#1274.9%#12013.6%#9933.1%#8640.1%#79ARC Prize

Coding

Benchmarkno reasoningminimallowmediumhighxhighdefaultSourceTrend
SciCode48.7%#68SciCode
LMArena Coding1477#94LMArena
LMArena WebDev1406#67LMArena
SciCode (AA)not in index49.6%#72Artificial Analysis
LiveCodeBench85.9%#25Vals AI
SWE-bench (Vals)not in index82.2%#20Vals AI

Agents & tools

Benchmarkno reasoningminimallowmediumhighxhighdefaultSourceTrend
LMArena Agent-6.9#36LMArena
Terminal-Bench 2.1 (Vals)55.1%#39Vals AI
MCP Atlas79.2%#10Scale AI SEAL

Maths

Benchmarkno reasoningminimallowmediumhighxhighdefaultSourceTrend
FrontierMath Tiers 1–346.3%#41Epoch AI Benchmarking Hub
FrontierMath Tier 417.1%#42Epoch AI Benchmarking Hub
OTIS Mock AIME90.0%#59Epoch AI Benchmarking Hub
ProofBench6.0%#53Vals AI

Knowledge

Benchmarkno reasoningminimallowmediumhighxhighdefaultSourceTrend
SimpleQA Verified19.1%#64Epoch AI Benchmarking Hub
AA-Omniscience-8.9#132Artificial Analysis
MMLU-Pro85.6%#56Vals AI
LegalBench83.0%#55Vals AI
CorpFin69.6%#6Vals AI
TaxEval75.5%#18Vals AI

Human preference

Benchmarkno reasoningminimallowmediumhighxhighdefaultSourceTrend
LMArena Text1407#125LMArena

Multimodal

Benchmarkno reasoningminimallowmediumhighxhighdefaultSourceTrend
LMArena Vision1234#60LMArena
MMMU-Pro74.0%#88Artificial Analysis

Long context

Benchmarkno reasoningminimallowmediumhighxhighdefaultSourceTrend
AA-LCR75.7%#107Artificial Analysis

Composite

Benchmarkno reasoningminimallowmediumhighxhighdefaultSourceTrend
Epoch Capabilities Indexnot in index150.2#42Epoch AI Benchmarking Hub
AA Intelligence Index26.1#110Artificial Analysis
Vals Indexnot in index31.9#41Vals AI

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

ProviderSpeedFirst tokenInput $/MOutput $/MContextQuantisation
Baseten180 tok/s0.27 s$0.500$1.201.0Mfp8
Together94 tok/s0.33 s$0.500$1.20524k
DeepInfra61 tok/s0.49 s$0.450$1.20524kfp8

Price history

Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.

$0.00$0.396$0.792$1.19$1.58Jul 26Aug 26
input outputnow $0.500 in · $1.20 out

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.100 per 1M. Reasoning tokens are not modelled.

WorkloadTokens in / outCostWith cachingTime
Chat reply400 / 300$0.0005$0.000417.5 s
Summarise a 30-page report12,000 / 600$0.0061$0.003019.5 s
Code edit6,000 / 1,500$0.0045$0.002925.5 s
Agentic coding session60,000 / 4,000$0.032$0.01642.2 s
Structured extraction2,000 / 200$0.0011$0.000616.8 s

See also

Data as of 9 Sept 2026. Compare these configurations.