tinyanalysis

Compare / LFM2.5 1.2B Thinking vs MiniCPM5 1B

LFM2.5 1.2B Thinking vs MiniCPM5 1B

Updated November 2026

Verdict

LFM2.5 1.2B Thinking scores higher on the Small LM Index: 34.7 vs 29.5, although their likely ranges overlap, so the lead is not certain. On the 5 benchmarks both labs report, LFM2.5 1.2B Thinking wins 1 and MiniCPM5 1B wins 4. Both need about 1.4 GB of memory at 4-bit quantization. Neither model has a hosted API on OpenRouter, so you would run both yourself.

Pick LFM2.5 1.2B Thinking if…

  • you want the strongest model overall

Pick MiniCPM5 1B if…

  • you need precise instruction following
  • you build agents that call tools
  • you work with long documents
  • you need a permissive license (Apache-2.0)
  • you want the more widely used model (more quantized builds and community support)

Specifications

LFM2.5 1.2B ThinkingMiniCPM5 1B
Small LM Index34.7 (likely 32–38)29.5 (likely 20–42)
LabLiquid AI · United StatesOpenBMB (MiniCPM) · China
Released2026-01-202026-05-21
Parameters1.2B (dense)1.1B (dense)
Memory at 4-bit~1.3 GB~1.4 GB
Context window32K tokens131K tokens
Inputtexttext
ReasoningAlways thinksHybrid (thinking on/off)
LicenseLFM Open LicenseApache-2.0
API price / 1M tokensSelf-host onlySelf-host only
Hugging Face downloads (30 days)8.3K257.9K

Benchmarks

Higher is better. Bold marks the winner where both labs report a score. ≈ marks the index model's estimate for a score that wasn't reported, ✓ an independent measurement and * a score measured by another lab.

BenchmarkLFM2.5 1.2B ThinkingMiniCPM5 1B
MMLU-Pro Knowledge49.7≈ 53
GPQA Diamond Knowledge33.9✓27.8✓
Humanity's Last Exam Reasoning6.2✓6.5✓
AIME Math31.7≈ 21
τ²-Bench Agentic19.6✓81.0✓
IFBench Instruction following41.8✓49.3✓
IFEval Instruction following88.4—
BFCL Agentic57.0—
MATH-500 Math88.0—
AA-LCR Long context0.0✓6.0✓

Scores are mostly reported by each lab with its own settings, so small gaps are noise. See the methodology. Open in the interactive comparison to add more models.

Questions

Which is better, LFM2.5 1.2B Thinking or MiniCPM5 1B?

LFM2.5 1.2B Thinking scores higher on the Small LM Index: 34.7 vs 29.5, although their likely ranges overlap, so the lead is not certain. On the 5 benchmarks both labs report, LFM2.5 1.2B Thinking wins 1 and MiniCPM5 1B wins 4.

Which is cheaper to run?

Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 1.4 GB of memory at 4-bit quantization.

Which has the longer context window?

MiniCPM5 1B has the longer context window: 131K tokens against 32K.

Can I use LFM2.5 1.2B Thinking and MiniCPM5 1B commercially?

LFM2.5 1.2B Thinking is released under LFM Open License and MiniCPM5 1B under Apache-2.0. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.

Related comparisons