LFM2.5-2.6B vs MiniCPM5-2B
Wins 0 of 6 areas
—
Wins 5 of 6 areas
Coding · Agents · Reasoning · Long documents · Following instructions
MiniCPM5-2B is the stronger all-rounder.
Scores updated · 29 tests both models report · How we compare
Where each one wins
Tests won in each of the six areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software03MiniCPM5-2B3 of 3 tests
- AgentsCarrying out multi-step tasks on its own03MiniCPM5-2B3 of 3 tests
- ReasoningHard problems that need careful thinking02MiniCPM5-2B2 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts01MiniCPM5-2B1 of 1 test
- Following instructionsDoing exactly what it is asked01MiniCPM5-2B1 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
Long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where MiniCPM5-2B pulls ahead
- Reasons across sets of long documentsAA-LCR+53.6points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+14.4points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+13.6points ahead
Where LFM2.5-2.6B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+5.9points ahead
Every test, side by side
All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingMiniCPM5-2B
- SciCodeMiniCPM5-2B by 11.914.426.3+11.9
- LiveCodeBench v6MiniCPM5-2B by 11.159.470.5+11.1
- Terminal-Bench 2.1MiniCPM5-2B by 4.14.58.6+4.1
AgentsMiniCPM5-2B
- τ-Bench V3 · BankingMiniCPM5-2B by 13.67.220.8+13.6
- GDPValMiniCPM5-2B by 10.4010.4+10.4
- MCP AtlasMiniCPM5-2B by 9.916.426.4+9.9
ReasoningMiniCPM5-2B
- GPQA DiamondMiniCPM5-2B by 14.455.870.2+14.4
- Humanity's Last ExamMiniCPM5-2B by 2.76.28.9+2.7
- CritPttie00.3tie
FactsEven
- AA-Omniscience · Non-hallucinationLFM2.5-2.6B by 5.98478.1+5.9
- AA-Omniscience · AccuracyMiniCPM5-2B by 44.38.4+4
Following instructionsMiniCPM5-2B
- IFBenchMiniCPM5-2B by 2.959.262.1+2.9
Other results16 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- τ³-bench (4 domains), averageMiniCPM5-2B by 12.656.869.4+12.6
- LiveCodeBench Pro 25Q2MiniCPM5-2B by 10.119.229.3+10.1
- BrowseComp-Plus (OpenClaw subset)LFM2.5-2.6B by 9.526.316.8+9.5
- SWE-bench Multilingual (OpenCode)MiniCPM5-2B by 7.7815.7+7.7
- IFEvalLFM2.5-2.6B by 7.695.487.8+7.6
- Artificial Analysis Coding IndexMiniCPM5-2B by 6.87.714.5+6.8
- AA Agentic IndexMiniCPM5-2B by 6.62.49+6.6
- WebArena-VerifiedMiniCPM5-2B by 6.139.1+6.1
- BFCLv4MiniCPM5-2B by 5.856.962.6+5.8
- Terminal-Bench 2.1 (OpenCode)MiniCPM5-2B by 5.605.6+5.6
- EnterpriseOps-GymMiniCPM5-2B by 5.57.913.4+5.5
- AA IntelligenceMiniCPM5-2B by 4.18.412.5+4.1
- Claw-Eval General (EN), pass^3LFM2.5-2.6B by 2.937.534.6+2.9
- Claw-Eval General (EN), averageLFM2.5-2.6B by 1.962.660.7+1.9
- HLE (with tools)MiniCPM5-2B by 1.99.611.5+1.9
- AA-Omnisciencetie-10.9-11.7tie
Questions people ask
Which is better, LFM2.5-2.6B or MiniCPM5-2B?
MiniCPM5-2B wins five of the six areas where both have results: coding, agents, reasoning, long documents and following instructions. LFM2.5-2.6B wins none. They are level on facts.
Which is better for coding?
MiniCPM5-2B. It wins 3 of the 3 coding tests both models report; LFM2.5-2.6B wins none.
How do you compare the two?
We use the 29 benchmark tests both models have published scores on. The verdict counts the 13 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 16 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.