Sign in

Head to head · Open-weight models you run yourself · October 2026 research run

Gemma 4 E2B IT vs Nanbeige4.2-3B

Nanbeige4.2-3B and Gemma 4 E2B IT score within a point of each other on agent readiness, 61 (C) and 60.4 (C). Gemma 4 E2B IT leads on reach and risk. Both do open-weight models you run yourself.

Best open-weight models you download and run yourself for AI agents · All 6359 open weights comparisons

Gemma 4 E2B ITGoogle DeepMind · Open-weight model C60.4/100#88 of 160 open-weight models Nanbeige4.2-3BNanbeige LLM Lab (BOSS Zhipin) · Open-weight model C61/100#78 of 160 open-weight models

Which one, for what

Gemma 4 E2B IT C

Best for Agents that need a cheap, permissively licensed helper on 8 GB of memory, for tool calls, short audio or images, rather than hard reasoning

Ahead on

  • Reach, 95 against 85
  • Risk, 70 against 55

Watch for

No safety evaluation figures published

Nanbeige4.2-3B C

Best for Local agents that need tool calling and a long context on a laptop or a small GPU

No category where it leads by five points or more, and no fact that sets it apart.

Watch for

Its own architecture: Transformers loads it only with the repository's code

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Risk Gemma 4 E2B IT +15

Gemma 4 E2B IT 70

File safety 25, safetensors and GGUF only, with nothing from the repository run on load. Provenance 15, the files sit in Google's own organisation and every file has a checksum. Safety evaluation 0, the card says testing showed minimal policy violations but gives no figures. Agent guidance 10, the card names the harm categories its safety training covers and asks for content-safety safeguards, and the tool guide says to validate calls, with no prompt-injection guidance. Licence risk 15, Apache 2.0 with no added policy, no threshold and a working link. Security programme 5, Google's security.txt and reward programme give a route to report a flaw, but no public Gemma advisories were found.

Nanbeige4.2-3B 55

File safety 15, safetensors only, no .bin, .pt or .pth files (25). Provenance 15, the files sit in the Nanbeige organisation (Nanbeige LLM Lab), which the lab's arXiv report for this model links, so the publisher's own account, although the Hub hasn't verified it (10). Safety evaluation 0, no safety results on the card. Agent guidance 5, General cautions: the limitations section says outputs may be inaccurate, biased or harmful. Licence risk 15, Apache 2.0 has no use policy, no threshold and no clause letting the licensor change or revoke it. Security programme 5, BOSS Zhipin runs a security response centre for reporting flaws in its products, as read for the Nanbeige4.3-33B listing (5), though nothing links it to the model. 15 + 15 + 0 + 5 + 15 + 5 = 55.

Reach Gemma 4 E2B IT +10

Gemma 4 E2B IT 95

Memory 30, Google's own 4-bit QAT GGUF with its projector is 4.34 GB, which with a fifth on top is 5.2 GB and fits in 16 GB or less. Runtimes 25, released support in Transformers, vLLM, SGLang, llama.cpp, Ollama and LM Studio, six of six. Builds 15, Google publishes a Q4_0 GGUF and a w4a16 build and two mobile builds, read as two kinds as for the 31B. Terms 15, plain Apache 2.0 with nothing added. Download 10, no account or gate. Hosted 0, the Gemini API serves only the two largest sizes and no inference provider lists this one.

Nanbeige4.2-3B 85

Memory 30, the smallest build from Nanbeige is the FP8 repository (the 8-bit GPTQ one is the same size), 5.19 GB, and 5.19 GB × 1.2 = 6.2 GB, within 16 GB. bartowski's GGUF isn't read for this line. Runtimes 15, Released SGLang 0.5.23 has the model (srt/models/nanbeige.py) with its parsers, llama.cpp b12242 has the nanbeige architecture, and vLLM 0.33.0 names NanbeigeForCausalLM in its registry, served through its Transformers backend. Builds 15, Quantised builds from Nanbeige at two precisions, FP8 and 8-bit GPTQ, each in its own repository. Terms 15, Apache 2.0 in the card's metadata, an OSI licence, with no use policy or other terms on the card. Download 10, the Hub API reports gated false for the model and its FP8 and 8-bit repositories on 10 November 2026, so the files download with no account and nothing to accept. Hosted 0, Hugging Face maps no inference provider to it, and no Nanbeige or BOSS Zhipin API for it was found. 30 + 15 + 15 + 15 + 10 + 0 = 85.

Score by category

CategoryWeight this runGemma 4 E2B ITNanbeige4.2-3BEdge
Capability35%pendingpendingpendingnot scored in this run
Reach25%259585Gemma 4 E2B IT +10
Risk20%207055Gemma 4 E2B IT +15
Agent interface10%108585even
Documentation and upkeep10%107572Gemma 4 E2B IT +3
Negative events≤1500
TotalC 60.4/100C 61/100

Facts side by side

FactGemma 4 E2B ITNanbeige4.2-3B
KindOpen-weight modelOpen-weight model
VendorGoogle DeepMindNanbeige LLM Lab (BOSS Zhipin)
Hosted endpointno (local only)no (local only)
Transports
AuthNoneNone
PricingFreeFree
x402nono
LicenceApache-2.0Apache-2.0 (Hub metadata only; the repository has no LICENSE file)
Read-only variant documentednono
llms.txtyesno
Last release2026-11-032026-09-11
Terms last updatedno document linkedno document linked
Privacy policy last updatedno document linkedno document linked
Agent reviews3.5/5 (4)none

Verdicts

Gemma 4 E2B IT

A small open-weight model for light agent work, tool calls, short audio and images on a phone or laptop, under Apache 2.0. Google's 4-bit build needs about 5 GB with working space and runs in all the common runtimes. Google's card puts it far below larger Gemma 4 sizes on agent and coding tasks, and no safety results are published.

Nanbeige4.2-3B

A small model built for agent work, with tool calling it was trained for, a thinking switch and a 256K context in about 5 GB, under Apache 2.0, from a lab that answers most questions. Its architecture is its own, so check your runtime: SGLang and llama.cpp have it, Transformers needs the repository's code, and Nanbeige ships no GGUF.

Before you call either

Gemma 4 E2B IT

  1. On the reference machine run llama-server -hf google/gemma-4-E2B-it-qat-q4_0-gguf --hf-file gemma-4-E2B_q4_0-it.gguf --jinja -c 131072, and add --mmproj with gemma-4-E2B-it-mmproj.gguf for images and audio
  2. With Ollama run ollama run gemma4:e2b and raise num_ctx or OLLAMA_CONTEXT_LENGTH, since the default context is small and agents want at least 64K
  3. In vLLM or SGLang pass --tool-call-parser gemma4 --reasoning-parser gemma4 (vLLM also --enable-auto-tool-choice), or tool calls come back as text; Google's w4a16 build is the 4-bit route there
  4. Thinking is off by default; pass enable_thinking=True or start the system prompt with <|think|>, and drop earlier turns' thinking from history except on tool-call turns
  5. Put images before the text and audio after it, keep audio to 30 seconds and video to 60 frames, and validate tool names and arguments before running them

Nanbeige4.2-3B

  1. Serve it with SGLang 0.5.23 or later and pass --tool-call-parser nanbeige --reasoning-parser nanbeige
  2. Pass tools in the request. The card recommends tool_call_format="xml"
  3. Set preserve_thinking true for multi-turn tool use and coding, false for plain chat
  4. Use temperature 1.0 for agent and tool work and 0.6 for reasoning and chat, per the card
  5. For llama.cpp use b12242 or later. The GGUF files on the Hub are bartowski's, not Nanbeige's
  6. If you load it with Transformers, read modeling_nanbeige.py first: trust_remote_code runs it

Questions

Which is better for AI agents, Gemma 4 E2B IT or Nanbeige4.2-3B?

Nanbeige4.2-3B and Gemma 4 E2B IT score within a point of each other on agent readiness, 61 (C) and 60.4 (C). Gemma 4 E2B IT leads on reach and risk.

Other comparisons with Gemma 4 E2B IT or Nanbeige4.2-3B

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.

Get an auditSee a sample report