LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
MiniCPM5 1B SFT
◆ OPEN
by OpenBMB
OPEN WEIGHTSLOCALSMALL
RANKED #196 OF 316 ASSISTANTS · OVERALL #294 OF 493 RANKED · VIBE SCORE 5.9 · 99 VOICES ·
COPY BADGE pastes a README snippet with the rank SVG.
MiniCPM5-1B-SFT on Hugging Face (text generation). 14,458 downloads. Open weights for local or hosted use.
HAPPY9.8
4 mentions
SAD5.2
3 mentions
VIBE TRAJECTORY
26 weeks · 106 voices in this window
HONEYMOON FADING
Early praise is cooling off in recent weeks.
HOW TO USE IT
Open weights. You can use a hosted service, or download it and run it yourself, free.
RUN IT ON YOUR OWN MACHINE
Open weights: download MiniCPM5 1B SFT once and it's yours. No subscription, no rate limits, works offline.
1.08B parameters, a Q4 file is about 660 MB (BF16 is ~2.1 GB)
- BARE MINIMUM
Raspberry Pi 5 (8GB)
8 GB UNIFIED RAM
~$80Q4_K_M, ~18 tok/s, 8k context via llama.cpp CPU inference
- SWEET SPOT
Apple Mac mini M2
8 GB UNIFIED MEMORY
~$450 USEDQ8_0 or FP16, ~85 tok/s, 32k context with MLX or llama.cpp Metal
- FULL POWER
NVIDIA GeForce RTX 3060 12GB
12 GB VRAM
~$280 USEDBF16 full precision, ~160 tok/s, full 131k context window in vLLM or SGLang
LEARN HOW TO SET IT UP
WILL MY BOX RUN MINICPM5 1B SFT?
The three cards above are researched picks. Search the machine you actually have — we only say yes if it should reply at a usable speed, not just load the weights and crawl.
HARDWARE PICKS & PRICES ARE RESEARCHED FROM THE LIVE WEB AND REFRESHED AUTOMATICALLY. TREAT THEM AS BALLPARK, NOT GOSPEL.
WHAT CAN YOU ACTUALLY DO WITH IT?
Everyday jobs on each of those rigs: how long MiniCPM5 1B SFT takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMRaspberry Pi 5 (8GB) | SWEET SPOTApple Mac mini M2 | FULL POWERNVIDIA GeForce RTX 3060 12GB |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~28 S★★★★★ | ~6 S★★★★★ | ~3 S★★★★★ |
| Summarize a documenta long report boiled down to the points that matter | ~83 S★★★★★ | ~18 S★★★★★ | ~9 S★★★★★ |
| Build a websitea small landing page, markup and styles together | ~7.4 MIN★★★★★ | ~1.6 MIN★★★★★ | ~50 S★★★★★ |
| Build a backendan API with routes, storage and tests | ~23 MIN★★★★★ | ~4.9 MIN★★★★★ | ~2.6 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~14 MIN★★★★★ | ~2.9 MIN★★★★★ | ~1.6 MIN★★★★★ |
STARS ARE THE COMMUNITY SCORE FOR THAT KIND OF WORK, DOCKED FOR HOW SQUASHED THE WEIGHTS GET AT EACH TIER. TIMES ARE COMPUTED FROM RESEARCHED SPEEDS. BALLPARK, NOT GOSPEL.
WHAT THE INTERNET SAYS
Not enough discussion about MiniCPM5 1B SFT yet to call it. We found 99 posts but people did not say much either way.
94 POSTS · 5 COMMENTS · GITHUB · REDDIT · X · HACKER NEWS · BLUESKY
PEOPLE LOVE
- No fan mail yet. The crowd is still warming up.
PEOPLE COMPLAIN ABOUT
- No complaints yet. The internet is still looking for one. Suspicious.
LOOKING FOR AN ALTERNATIVE?
Other models the crowd has fully reviewed, starting with text models like this one.
Muse Spark 1.3
DeepSeek V4 Flash> SOURCES (12)
- REDDIT · 2026-06-01 · 537 UPVOTESLocal AI News You Missed - May 2026
Local AI News You Missed - May 2026
- REDDIT · 2026-07-20 · 189 UPVOTESopenbmb released MiniCPM5-2B, not yet available at huggingface
openbmb released MiniCPM5-2B, not yet available at huggingface
- REDDIT · 2026-07-07 · 17 UPVOTEStraining times for a 1B model is too slow on my ai workstation with RTX PRO 6000
training times for a 1B model is too slow on my ai workstation with RTX PRO 6000
- REDDIT · 2026-06-15 · 5 UPVOTESI fine-tuned MiniCPM5-1B to turn Chinese astrology into gentle self-reflection — 100% local via llama.cpp, open dataset + scripts
I fine-tuned MiniCPM5-1B to turn Chinese astrology into gentle self-reflection — 100% local via llama.cpp, open dataset + scripts
- HACKER NEWS · 2026-09-07 · 4 UPVOTESMiniCPM5-2B – new leading ≤4B model
MiniCPM5-2B – new leading ≤4B model
- REDDIT · 2026-05-30 · 3 UPVOTESMinicpm5 1b is the first tiny model release that made me rethink the floor
Minicpm5 1b is the first tiny model release that made me rethink the floor
- GITHUB · 2026-10-05 · 2 UPVOTESfeat: add ollama-native extension with streaming and tool calling
feat: add ollama-native extension with streaming and tool calling
- GITHUB · 2026-10-06 · 1 UPVOTESllama-perplexity mis-scores GLM-4.7-Flash at -ub 1 (PPL 264.97 vs 10.74; generation is fine)
llama-perplexity mis-scores GLM-4.7-Flash at -ub 1 (PPL 264.97 vs 10.74; generation is fine)
- GITHUB · 2026-08-21 · 0 UPVOTESfeat(gpu): re-enable the Vulkan LLM lane behind an image gate
feat(gpu): re-enable the Vulkan LLM lane behind an image gate
- GITHUB · 2026-09-08 · 0 UPVOTES[Bug]: llama.cpp v0.4.0-dev 新版本号格式解析失败，has_update 误判"有更新"并静默回滚用户升级的运行时
[Bug]: llama.cpp v0.4.0-dev 新版本号格式解析失败，has_update 误判"有更新"并静默回滚用户升级的运行时
- GITHUB · 2026-09-14 · 0 UPVOTESfeat(amd): add DeepSeek-V4.1-Flash, Qwen3.8-27B, MiniCPM5-2B; drop retired MiniCPM5-1B
feat(amd): add DeepSeek-V4.1-Flash, Qwen3.8-27B, MiniCPM5-2B; drop retired MiniCPM5-1B
- GITHUB · 2026-09-26 · 0 UPVOTESZAYA1 (Zyphra) on the GPU: pin llama.cpp 5556bf2, serve routes fork-only architectures
ZAYA1 (Zyphra) on the GPU: pin llama.cpp 5556bf2, serve routes fork-only architectures