Skip to content
OSMEXO

LFM2.5-2.6B

liquidai/lfm-2.5-2.6b
  • Liquidai
  • Family · LFM2
  • Released 4 Aug 2026
  • Open weights
  • Text → Text
  • Tools · JSON
Not on OSMEXO yet
Spec & list price
Context125K
Max output8K
Input / 1M$0.35
Output / 1M$0.501.4× input
Scores · standing
Intelligence8.4#296/489
Coding7.7#183/196
Usage
Tokens / day1.3B▲ 189% 7d
Usage rank#183
Requests / day411
Tool calls / req0.21
Per request · 30 days
Avg input3.7K0.3× median
Avg output696about median
Cache hit30%0.4× median
Reasoning share80%1.4× median
Tool-call failures1.5%0.8× median

Consumption

3.3M6.7M10.0M3 Oct17 Oct1 Nov
Peak 8.7M 28 OctLow 47.6K 23 OctRank today #183

News & trends

Trending · 14 days · LFM2 family
News · latest from Liquidai

Benchmarks

bar = share of the top score · Benchmark data: Artificial Analysis
GPQA Diamond#296 of 445
55.8%
SciCode#159 of 160
14.4%
Intelligence index#296 of 489
8.4
τ-Bench Banking#113 of 154
7.2%
Humanity's Last Exam#283 of 465
6.2%
Coding index#183 of 196
7.7
Long-context reasoning#337 of 393
5.7%
Terminal-Bench 2.1#156 of 182
4.5%
Speed · median across 2 providers, latest day
Output speed#8/334266 tok/s4.0× faster than median
Fastest provider376 tok/sbest of 2 providers
First token#50.23 s4.5× quicker than median

#24 cheapest of 44 with a similar score

Value
Intelligence Index
68101214$0.2$0.5$1LFM2.5-2.6B
Blended price per 1M tokens (log)
Family progress
Intelligence Index
456789SepNov2026MarMayJulLFM2.5-2.6B
  • Family releases8.4

Technical details

Tokenizer
Other
License
other
Parameters
2.7B
Moderation
none
Downloads · likes
69.6K · 859
Request parameters 19
  • tools
  • tool_choice
  • response_format
  • structured_outputs
  • reasoning
  • include_reasoning
  • temperature
  • top_p
  • top_k
  • min_p
+9 more
  • max_tokens
  • max_completion_tokens
  • stop
  • seed
  • frequency_penalty
  • logprobs
  • presence_penalty
  • repetition_penalty
  • top_logprobs

Interesting facts

  1. Less output room than similar models (8.2K against a 49.2K median).
  2. Released 90 days ago, the newest of 8 LFM2 models.
  3. Weekdays average 837.7K tokens a day against 874.0K on weekends.

When to use

Best for

  • On-device and edge assistants
  • Lightweight classification and routing
  • Cheap, fast text rewriting and extraction
  • Self-hosted deployments (open weights)

Think twice for

  • Coding agents
  • Long document Q&A (small context)
  • Complex multi-step reasoning
  • Image input (text only)

Same family

LFM2

Alternatives

other vendors · similar capability

Want LFM2.5-2.6B on OSMEXO?

Sign in to request it. Requests are reviewed; a request does not guarantee availability.