GoogleLanguage model
Gemma 4 E2B IT
Gemma 4 E2B IT has no overall index value yet because there are too few independent benchmark results. From the vendor-reported results we estimate about 34.0 points (95% range 11.2 to 56.7). This is not an independent measurement.
58out of 100+12 vs. last month
based on search, usage and attention
Summary
What is new in Google Gemma 4 E2B IT is the smallest instruction-tuned Gemma 4 variant, with 2.3 billion effective parameters and 5.1 billion including embeddings. That distinction explains its local-device focus without making E2B a 2B total model.
Apache 2.0 and the compact parameter size mark the checkpoint’s on-device direction. The E stands for effective parameters, not a shortened model name. Gemma 4 E4B IT offers more capacity while keeping the same on-device direction.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 101 published benchmark results
Page 1 of 9
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
MMLU Pro (Show measurement, test conditions, and source)
- Source value
- 60
- Score
- 60
- Metric
- percent
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google Gemma 4 official model card
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Gemma 4 E2B instruction-tuned column; card does not state per-task prompts, decoding setup, or evaluation date.
AIME 2026 no tools (Show measurement, test conditions, and source)
- Source value
- 37.5
- Score
- 37.5
- Metric
- percent
- Unit
- %
- Benchmark version
- AIME 2026
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google Gemma 4 official model card
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Gemma 4 E2B instruction-tuned; explicitly labeled no tools; other prompting and decoding details are not stated.
LiveCodeBench v6 (Show measurement, test conditions, and source)
- Source value
- 44
- Score
- 44
- Metric
- percent
- Unit
- %
- Benchmark version
- LiveCodeBench v6
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google Gemma 4 official model card
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Gemma 4 E2B instruction-tuned; task interval and decoding settings are not disclosed.
GPQA Diamond (Show measurement, test conditions, and source)
- Source value
- 43.4
- Score
- 43.4
- Metric
- percent
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google Gemma 4 official model card
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Gemma 4 E2B instruction-tuned; prompt and decoding details are not stated.
MMMLU (Show measurement, test conditions, and source)
- Source value
- 67.4
- Score
- 67.4
- Metric
- percent
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google Gemma 4 official model card
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Gemma 4 E2B instruction-tuned; the model card does not state task prompting or decoding parameters.
MMMU Pro (Show measurement, test conditions, and source)
- Source value
- 44.2
- Score
- 44.2
- Metric
- percent
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google Gemma 4 official model card
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Gemma 4 E2B instruction-tuned; split and decoding details are not stated.
MATH-Vision (Show measurement, test conditions, and source)
- Source value
- 52.4
- Score
- 52.4
- Metric
- percent
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google Gemma 4 official model card
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Gemma 4 E2B instruction-tuned; task split and decoding details are not stated.
MRCR v2 8-needle 128K average (Show measurement, test conditions, and source)
- Source value
- 19.1
- Score
- 19.1
- Metric
- percent
- Unit
- %
- Benchmark version
- MRCR v2; 8 needles; 128K
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google Gemma 4 official model card
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Gemma 4 E2B instruction-tuned; card labels the average score but omits detailed run parameters.
Tau2 (Show measurement, test conditions, and source)
- Source value
- 24.5
- Score
- 24.5
- Metric
- percent
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Gemma 4 model card
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Gemma 4 E2B column in the Google Gemma 4 model card, final IT model. The table labels this an average over three tasks.
- Context
- Model-card result; no per-result sample count or confidence interval is stated. The source does not identify a more specific score metric. The source does not state a benchmark-suite version.
BigBench Extra Hard (Show measurement, test conditions, and source)
- Source value
- 21.9
- Score
- 21.9
- Metric
- percent
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Gemma 4 model card
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Gemma 4 E2B column in the Google Gemma 4 model card, final IT model.
- Context
- Model-card result; no per-result sample count or confidence interval is stated. The source does not identify a more specific score metric. The source does not state a benchmark-suite version.
MedXPertQA MM (Show measurement, test conditions, and source)
- Source value
- 23.5
- Score
- 23.5
- Metric
- percent
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Gemma 4 model card
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Exact Gemma 4 E2B column in the Google Gemma 4 model card, final IT model.
- Context
- Model-card result; no per-result sample count or confidence interval is stated. The source does not identify a more specific score metric. The source does not state a benchmark-suite version.
T1-Bench: Tool call accuracy (Show measurement, test conditions, and source)
- Source value
- 48.89
- Score
- 48.89
- Metric
- Tool Call > Acc.
- Unit
- %
- Benchmark version
- T1-Bench arXiv:2606.11070v1, Table 2 (overall, all domains)
- Category
- agentic-workflow
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- T1-Bench paper (arXiv:2606.11070v1)
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Target row: Gemma4-E2B-it. T1-Bench Table 2 reports overall results averaged over all domains. The paper states that all user and judge agents are instantiated as Gemma4-26B-A4B-it. Table 6 lists temperature=1.0, top_p=0.95, top_k=64 and Reasoning Effort '-'. The served model route and immutable evaluated-checkpoint revision are not reported.
- Context
- The paper reports Table 2 aggregates averaged across domains but does not provide the model-row denominator or a confidence interval for these aggregate cells. These are tool-use task-system metrics, not bare-model output-quality scores. Averaging across domains does not supply an instance-level N. The served route and immutable evaluated-checkpoint revision are not identified.
Page 1 of 9
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
The 100 most popular models of this category, 0 to 100, higher is more popular
20 of 100 models
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 7, 2026
Gemma 4 E2B ITGPT-5.4 mini 59Qwen3.5 2B 58DeepSeek-R1 58Phi-4 57
Method changes in the timeline, a jump at these points is no market event:
- 1November 1, 2026: Value only with search or usage data, 30-day attention
What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.
Similar models
More models of the same kind as Gemma 4 E2B IT, chosen by model line, developer, size, price and release date.
- Gemma 4 E4B ITGoogle · same developer
- Gemma 3n E2B ITGoogle · same developer
- Gemma 3 4BGoogle · same developer
- Gemma 3n E4B ITGoogle · same developer
- Gemma 4 12B ITGoogle · same developer
- Gemma 4 26B A4BGoogle · same developer
- DiffusionGemma 26B A4B ITGoogle · same developer
- Gemma 4 31BGoogle · same developer
All specifications
Architecture
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
Frequently asked questions about Gemma 4 E2B IT
How good is Gemma 4 E2B IT in benchmarks?
Independent results are not yet enough for the overall index. An estimate from 3 vendor-reported results (GPQA Diamond 43.4%, LiveCodeBench v6 44%, MMLU-Pro 60%) gives about 34.0 points, with a 95% range of 11.2 to 56.7. The estimate is calibrated on models with both vendor-reported and independent results. Vendor numbers run clearly above independent measurements on some tests; the calibration offsets that on average.
Vendor estimate, not an overall index value
How large is the context window of Gemma 4 E2B IT?
131,072 tokens. That is how much text Gemma 4 E2B IT reads at once, about 200 A4 pages.
Page count as a rule of thumb from Artificial Analysis
Can I run Gemma 4 E2B IT locally?
Yes, the weights are open.
Stated by Google
- Open weights
- yes
- License
- Apache 2.0
- Parameters
- 5.1B
- active per token
- n/a
Sources and data date
Show all 33 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. Retrieved October 27, 2026. |
| Output | Documented modalities. Retrieved October 27, 2026. |
| Research status | Research date October 15, 2026. 26 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not published | benchmark-protocol-details Note The card does not specify per-task prompts, decoding settings, or evaluation dates. |
| Not published | immutable model revision Original wording The pinned evaluation config explicitly leaves each Hugging Face `revision` null; no immutable weights revision can be established from this source. |
| Not published | model-side reasoning or effort setting Original wording The paper identifies prompt strategy but does not publish a separate model-side thinking/effort toggle for these exact conditions. |
| Additional source | Hugging Face (retrieved October 15, 2026) · Gemma 4 E2B instruction-tuned weights · Editorial description reviewed October 2, 2026 |
| Additional source | Gemma 4 model card (retrieved October 17, 2026) |
| Additional source | Gemma 4 Technical Report v2 (retrieved October 17, 2026; October 18, 2026) |
| Additional source | Unified Deployment-Aware Evaluation (UDAE), pinned public result CSV (retrieved October 17, 2026)12 |
| Additional source | Unified Deployment-Aware Evaluation (UDAE), TMLR v3, Table 16 (retrieved October 17, 2026) |
| Additional source | T1-Bench paper (arXiv:2606.11070v1) (retrieved October 17, 2026) |
| Additional source | Google DeepMind Gemma 4 family page (retrieved October 17, 2026) |
| Additional source | CatfishW Gemma 4 E2B/E4B Reasoning Evaluation README (retrieved October 18, 2026) |
| Additional source | OpenBMB MiniCPM5-2B model card (retrieved October 18, 2026) |
| Additional source | IDP Leaderboard, Gemma-4-E2B-it model page (retrieved October 20, 2026) |
| Additional source | AI Eval Corp, AIEval DoW Bench (retrieved October 21, 2026) |
| Additional source | Gemma Team, Google DeepMind (retrieved October 22, 2026) |
| Additional source | Model metadata (retrieved October 17, 2026)123456789101112 |
| Additional source | Model metadata (retrieved October 18, 2026) |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Ollama: cumulative downloads for the popularity index |
| Additional source | Replicate: cumulative runs for the popularity index |
| Additional source | Google Trending Now: limited feed selection for Germany and the US |
| Additional source | Creative Commons Attribution 4.0 |