BenchLeader
Allen Institute for AIOpen weightsReasoning modelAuto-detected

Olmo 3 7B Think

Best configuration ranks #520 of 610 on the BenchLeader Index at 39.8 ±5.6 (thinking reasoning effort).

Blended price
$0.125/M
$0.100 in · $0.200 out
Output speed
First answer
Context
66k
How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.
  1. Overall index40
  2. Agents & tools34
  3. Knowledge30
  4. Instruction following44
  5. Long context25
  6. Composite36

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
thinkingbest39.8#520$0.0001Agents & tools 34 · Composite 36 · Instruction following 44 · Knowledge 30 · Long context 25
default38.4#551$0.0001Agents & tools 34 · Composite 36 · Instruction following 36 · Knowledge 28 · Long context 25

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

BenchmarkthinkingdefaultSource
GPQA Diamond (AA)not in index51.6%#39740.0%#454Artificial Analysis
Humanity's Last Exam (AA)not in index6.0%#3345.8%#341Artificial Analysis

Agents & tools

BenchmarkthinkingdefaultSource
Terminal-Bench Hard0.8%#3320.0%#345Artificial Analysis
τ²-Bench Telecom (AA)not in index0.0%#37512.6%#358Artificial Analysis

Knowledge

BenchmarkthinkingdefaultSource
AA-Omniscience-71.8#442-77.1#452Artificial Analysis

Instruction following

BenchmarkthinkingdefaultSource
IFBench41.5%#24132.8%#321Artificial Analysis

Long context

BenchmarkthinkingdefaultSource
AA-LCR0.0%#4190.0%#419Artificial Analysis

Composite

BenchmarkthinkingdefaultSource
AA Intelligence Index5.6#4815.2#502Artificial Analysis

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. Reasoning tokens are not modelled.

WorkloadTokens in / outCostWith cachingTime
Chat reply400 / 300$0.0001
Summarise a 30-page report12,000 / 600$0.0013
Code edit6,000 / 1,500$0.0009
Agentic coding session60,000 / 4,000$0.0068
Structured extraction2,000 / 200$0.0002

See also

Data as of 9 Sept 2026. Compare these configurations.