MiniCPM5-2B
MiniCPM5-2B is behind the leaders in long context, instruction following, factuality, and agentic tasks. Too few results yet to rate reasoning, coding, safety, math, multimodal tasks, or multilingual tasks.
Capability profile
Bars show the model's median result as a share of the leading model's, per capability. Select a row to see its results.
Too few results yet to rate Reasoning, Coding, Safety, Math, Multimodal or Multilingual.
Price
No current price is tracked for this model. See the rate card
Evidence
36results on33benchmarks
- 15 aggregator
- 21 cross-referenced
From 2 sources · latest Oct 28, 2026 · How verification works
MiniCPM5-2B benchmark results
36 results on 33 benchmarks, grouped by capability. Each bar is the result as a share of the capability leader's; every result links to its source.
32.2% behind the leader1 of 3 ranked benchmarks measured
- 59.33Oct 16, 2026
33.0% behind the leader1 of 3 ranked benchmarks measured
- 62.08Oct 28, 2026
35.7% behind the leader2 of 4 ranked benchmarks measured
- AA-Omniscience · Non-hallucination78.09Oct 16, 2026omniscienceNonHallucination
- AA-Omniscience · Accuracy8.42Oct 16, 2026omniscienceAccuracy
49.5% behind the leader4 of 7 ranked benchmarks measured
- τ-Bench V3 · Banking20.82Oct 16, 2026tauBanking
- 26.38Oct 28, 2026
- 10.36Oct 16, 2026
- Terminal-Bench 4.00.00Oct 16, 2026
0 of 6 ranked benchmarks measured
- 0.29Oct 16, 2026
- GPQA Diamond70.20Oct 16, 2026gpqa
- Humanity's Last Exam8.94Oct 16, 2026aa_hle
Show 2 more reasoning resultsHide 2 reasoning results
- 65.15Oct 28, 2026
- Humanity's Last Exam8.20Oct 28, 2026Humanity's Last Exam (no tools)
0 of 10 ranked benchmarks measured
- SciCode26.27Oct 16, 2026aa_scicode
- Terminal-Bench 2.18.61Oct 16, 2026terminalbenchV21
- 27.78Oct 28, 2026
Show 2 more coding resultsHide 2 coding results
- 70.52Oct 28, 2026
- SWE-bench Verified34.80Oct 28, 2026SWE-bench Verified (OpenCode)
More results
Benchmarks outside the capability baskets. They are not ranked against a leader.
- AA Intelligence12.46Oct 16, 2026aa_intelligence_index
- -11.65Oct 16, 2026
- Artificial Analysis Coding Index14.50Sep 9, 2026aa_coding_index
- 8.95Sep 9, 2026
- 37.96Oct 28, 2026
- HLE (with tools)11.50Oct 28, 2026Humanity's Last Exam (with tools)
Show 12 more resultsHide 12 results
- AutomationBench Public4.71Oct 28, 2026AutomationBench (public)
- 62.65Oct 28, 2026
- BrowseComp-Plus (OpenClaw subset)16.83Oct 28, 2026
- Claw-Eval General (EN), average60.74Oct 28, 2026
- Claw-Eval General (EN), pass^334.58Oct 28, 2026
- EnterpriseOps-Gym13.41Oct 28, 2026
- 87.83Oct 28, 2026
- LiveCodeBench Pro 25Q229.34Oct 28, 2026
- SWE-bench Multilingual (OpenCode)15.67Oct 28, 2026
- Terminal-Bench 2.1 (OpenCode)5.62Oct 28, 2026
- 9.11Oct 28, 2026
- τ³-bench (4 domains), average69.44Oct 28, 2026
MiniCPM5-2B: common questions
Who makes MiniCPM5-2B?
MiniCPM5-2B is made by OpenBMB.
When was MiniCPM5-2B released?
MiniCPM5-2B was released on Sep 7, 2026, according to Artificial Analysis.
What is MiniCPM5-2B good at?
MiniCPM5-2B is behind the leaders in long context, instruction following, factuality, and agentic tasks. Too few results yet to rate reasoning, coding, safety, math, multimodal tasks, or multilingual tasks.
How many benchmarks has MiniCPM5-2B been tested on?
We track 36 results for MiniCPM5-2B on 33 benchmarks from 2 sources. The latest was recorded on Oct 28, 2026.
About this record
Where MiniCPM5-2B's numbers come from, and every name it appears under.
- Tracked since
- Sep 7, 2026
- Newest source mention
- Sep 28, 2026
Where the results come from
Verification: 36 scores · 0 independently verified · 15 aggregator-attributed · 21 vendor cross-reference. How these tiers are assigned
From 2 sources on 2 sites. Hugging Face supplies 21 of them. Bars are coloured by trust tier.
- huggingface.co21
- artificialanalysis.ai15