Skip to content
Model Pareto

LFM2.5-2.6B

Liquid AIOpen weightsartificialanalysis.ai ↗
Input
—
Output
—
Speed
~110tok/sest.
Context
128Ktokens

Long-tail import from Artificial Analysis (2026-09-24); AA variant: LFM2.5-2.6B. Price = AA median host price.

Rankings

Self-hostingEstimated

How we estimate →

What it would cost to run these open weights yourself on rented GPUs. No API sells this model, so this is its price on the chart.

Hardware
1× B200 180GB
BF16 weights
Throughput
~19,533 tok/s
many requests batched
Price
$0.037–0.33 /M tok
busy → light use
On your own machine
RTX 5090 32GB (int4, 300+ tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • BF16 weights (2.69B params) + 25% KV-cache headroom ≈ 7 GB
  • 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~19,533 output tok/s aggregate at batch 256 (bandwidth-bound)
  • Blended 3:1 input:output; prefill ~197,607 tok/s
  • Low = 75% utilization at the low GPU price; high = 20% at the high price

Benchmark scores

~ italic, dashed = no published score yet, estimated from related benchmarks.

Coding

~last of 21 (estimated)below measured range (<51.0%)
  1. #1Claude Opus 596.0%
  2. #2Claude Sonnet 585.2%
  3. #20LFM2.5-24B-A2B51.0%
  4. ~#21LFM2.5-2.6BEstimated<51.0%
No published score — estimated from related benchmarks
~last of 24 (estimated)below measured range (<43.8%)
  1. #1Claude Fable 5.293.0%
  2. #2Claude Opus 5.589.9%
  3. #23Nex-N2.5-Mini43.8%
  4. ~#24LFM2.5-2.6BEstimated<43.8%
No published score — estimated from related benchmarks
~#63 of 102 (estimated)~0.1%
  1. #1Kimi K3.1 Pro67.2%
  2. #2Claude Sonnet 5.563.6%
  3. #53Trinity Large Thinking0.5%
  4. ~#63LFM2.5-2.6BEstimated0.1%
  5. #63A.X-K20.0%
No published score — estimated from related benchmarks
LiveCodeBenchEstimated
~#25 of 28 (estimated)~32.4%
  1. #1Qwen3.8-Omni-Flash92.6%
  2. #2Qwen3.8 Flash-Next91.9%
  3. #24Devstral Small 234.8%
  4. ~#25LFM2.5-2.6BEstimated32.4%
  5. #25Ministral 3 8B30.3%
No published score — estimated from related benchmarks
~last of 26 (estimated)below measured range (<26.3)
  1. #1Claude Opus 5.561.7
  2. #2Claude Opus 559.7
  3. #25Muse Spark Open26.3
  4. ~#26LFM2.5-2.6BEstimated<26.3
No published score — estimated from related benchmarks

Hard reasoning

Agents

~#12 of 15 (estimated)~56.5%
  1. #1Kimi K384.8%
  2. #2Qwen3.8 27B84.3%
  3. #11Muse Glimmer65.9%
  4. ~#12LFM2.5-2.6BEstimated56.5%
  5. #12Nemotron 3.5 Super VL NVFP452.3%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~last of 5 (estimated)below measured range (<47.5%)
  1. #1GLM-5.3 Flash59.1%
  2. #2Kimi K358.3%
  3. #4GPT-5.547.5%
  4. ~#5LFM2.5-2.6BEstimated<47.5%
No published score — estimated from related benchmarks
BrowseCompEstimated
~#19 of 22 (estimated)~50.3%
  1. #1Atria Dawn Preview92.5%
  2. #2GPT-6 Astra91.5%
  3. #18Nemotron 3.5 Bolt 74B-A9B70.0%
  4. ~#19LFM2.5-2.6BEstimated50.3%
  5. #19Mistral Medium 3.548.6%
No published score — estimated from related benchmarks
~#66 of 92 (estimated)~536 Elo
  1. #1Claude Opus 5.51866
  2. #2Muse Wonder Full1857
  3. #65Gemma 4 26B A4B553
  4. ~#66LFM2.5-2.6BEstimated536
  5. #66Devstral 2531
No published score — estimated from related benchmarks
~#73 of 85 (estimated)~0.8%
  1. #1Claude Fable 5.276.4%
  2. #2Muse Wonder Full71.3%
  3. #72Nemotron 3.5 Lightning0.8%
  4. ~#73LFM2.5-2.6BEstimated0.8%
  5. #73Ministral 3 3B0.7%
No published score — estimated from related benchmarks