Gemma 4 26B A4B: benchmark scores, price and ranking
Gemma 4 26B A4B from Google DeepMind ranks #432 of 743 LLMs on the LLMs Tiger aggregate benchmark score (47.5), based on 11 benchmark results captured up to 2026-10-13.
- Organisation
- Google DeepMind
- Type
- Open-weight / local model
- Open weights
- Yes
- Released
- 2026-04-02 (source: epoch)
- Cheapest API price
- $0.076 in / $0.255 out per million tokens (via openrouter, observed 2026-10-13T19:47Z)
- Aggregate rank
- #432 of 743 (score 47.5)
Relative to other models, Gemma 4 26B A4B is strongest on Epoch ECI (#44 of 204) and weakest on CritPt (#90 of 115).
Across the 11 benchmarks where at least ten models are scored, Gemma 4 26B A4B is in the top 10% on 0 and in the top half on 6.
Its cheapest listed input price, $0.076 per million tokens, is lower than that of 87% of the 427 models with a listed price.
Gemma 4 26B A4B benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| Epoch ECI | 142 | #44 of 204 | Epoch AI Benchmarking Hub | 2026-10-13 |
| ALE-Bench | 927 | #26 of 79 | Epoch AI Benchmarking Hub | 2026-10-13 |
| OTIS Mock AIME | 82.2% | #63 of 189 | Epoch AI Benchmarking Hub | 2026-10-13 |
| DTBench | 74.9% | #67 of 146 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LMCA | 29.7% | #51 of 108 | Epoch AI Benchmarking Hub | 2026-10-13 |
| GPQA Diamond | 73.2% | #135 of 275 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Chess Puzzles | 6.0% | #73 of 122 | Epoch AI Benchmarking Hub | 2026-10-13 |
| SciCode | 40.0% | #66 of 106 | Epoch AI Benchmarking Hub | 2026-10-13 |
| WebDev Arena | 1,361 | #44 of 67 | Epoch AI Benchmarking Hub | 2026-10-13 |
| WeirdML | 35.2% | #80 of 109 | Epoch AI Benchmarking Hub | 2026-10-13 |
| CritPt | 0.0% | #90 of 115 | Epoch AI Benchmarking Hub | 2026-10-13 |
About Gemma 4 26B A4B
Gemma 4 26B A4B is an open-weight model that can run on your own hardware from Google DeepMind. It was released on 2026-04-02, according to Epoch AI. The cheapest API listing we track charges $0.076 per million input tokens and $0.255 per million output tokens, through openrouter. We hold 11 benchmark results for it across 4 areas, and none is in the top tenth of its board.
Where Gemma 4 26B A4B stands
On our aggregate score, Gemma 4 26B A4B is #432 of 743 ranked models, around the middle of the table. Its score is 47.5; the scale is centred on 50, so it sits 2.5 points below the typical model. That beats 42% of the ranked field. Among the 41 ranked Google DeepMind models it is #21. The next one up is gemini-1.5-pro-001-feb24 at 48.1. The next one down is Gemma 4 31B IT at 47.3.
Gemma 4 26B A4B, benchmark by benchmark
Epoch ECI puts Gemma 4 26B A4B at 142, #44 of 204. That beats 78% of the models tested, but 15 points short of Qwen 3.8 Max (156). The results either side are Qwen3-Max (142) and Gemini 2.5 Flash (May 2025) (142). It covers Epoch Capabilities Index: Epoch's own aggregate of ~40 benchmarks. The board median is 127, so Gemma 4 26B A4B is 15 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On ALE-Bench, Gemma 4 26B A4B reaches 927. It ranks #26 of 79, ahead of 67% of the field, and 2,024 points behind GPT-6 Astra (max) at 2,951. The results either side are o3 (high) (934) and Gemma 4 31B IT (926). It covers AtCoder heuristic-optimisation contests (performance rating). The board median is 764, so Gemma 4 26B A4B is 163 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Gemma 4 26B A4B manages 82.2% on OTIS Mock AIME, good for #63 of 189. Claude Fable 5.1 (max) leads at 100.0%, 17.8 points further up, and 67% of models sit below. The results either side are Kimi K2 Thinking Turbo (83.1%) and o4-mini (high) (81.7%). It covers competition maths (AIME 2024-25 style). The board median is 57.8%, so Gemma 4 26B A4B is 24.4 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
DTBench puts Gemma 4 26B A4B at 74.9%, #67 of 146. That beats 54% of the models tested, but 24.0 points short of Claude Opus 5.5 (max) (98.9%). The results either side are Mistral Medium 3.5 (75.5%) and o1 (high) (74.7%). It covers dynamic tool-use benchmark. The board median is 68.8%, so Gemma 4 26B A4B is 6.1 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On LMCA, Gemma 4 26B A4B reaches 29.7%. It ranks #51 of 108, ahead of 53% of the field, and 38.6 points behind Claude Opus 5.5 (max) at 68.2%. The results either side are Qwen3.6 35B-A3B (29.7%) and mimo-v2.5-pro (29.5%). It covers language-model capability assessment. The board median is 29.2%, so Gemma 4 26B A4B is 0.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Gemma 4 26B A4B manages 73.2% on GPQA Diamond, good for #135 of 275. GPT-6 Astra (max) leads at 95.8%, 22.5 points further up, and 51% of models sit below. The results either side are claude-opus-4-1-20250805 (73.2%) and Qwen3-Max-Instruct (72.6%). It covers expert-level science questions. The board median is 71.7%, so Gemma 4 26B A4B is 1.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Gemma 4 26B A4B lands below the middle on SciCode: 40.0%, #66 of 106, with 61% of the field scoring higher. Claude Opus 5.5 (max) leads at 66.9%; the closest result above is Claude Sonnet 4 (unknown thinking) at 40.0%. It covers scientific research code. The board median is 44.3%, so Gemma 4 26B A4B is 4.3 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Gemma 4 26B A4B lands below the middle on WebDev Arena: 1,361, #44 of 67, with 64% of the field scoring higher. Claude Opus 5.5 (max) leads at 1,820; the closest result above is Gemma 4 31B IT at 1,364. It covers building web apps, crowd-voted (Elo). The board median is 1,433, so Gemma 4 26B A4B is 72 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Gemma 4 26B A4B lands below the middle on WeirdML: 35.2%, #80 of 109, with 72% of the field scoring higher. GPT-6 Astra (max) leads at 93.3%; the closest result above is gemini-2.5-flash-lite-preview-06-17_16K at 35.2%. It covers unusual machine-learning tasks. The board median is 41.2%, so Gemma 4 26B A4B is 6.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Gemma 4 26B A4B lands below the middle on CritPt: 0.0%, #90 of 115, with 77% of the field scoring higher. GPT-5.6 Sol (max) leads at 32.3%; the closest result above is Devstral Small 2 at 0.0%. It covers unpublished research-level physics problems. The board median is 2.9%, so Gemma 4 26B A4B is 2.9 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
The full table above also lists the other 1 result, including boards with fewer than ten models, where a rank says little.
Strengths and weaknesses
Software engineering: uneven, ahead of 46% of the field on average across 3 benchmarks. Its best board here is ALE-Bench (#26 of 79) and its weakest is WebDev Arena (#44 of 67). Reasoning, maths and science: uneven, ahead of 46% of the field on average across 3 benchmarks. Its best board here is OTIS Mock AIME (#63 of 189) and its weakest is CritPt (#90 of 115). General knowledge and chat: uneven, ahead of 49% of the field on average across 4 benchmarks. Its best board here is Epoch ECI (#44 of 204) and its weakest is WeirdML (#80 of 109).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so Gemma 4 26B A4B costs less than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on qwen3-4b ($0.080) and Qwen3-32B ($0.080).
Release history
The Google DeepMind model released before it was Gemini 3 Deep Think on 2026-02-12, which scores 52.6. The next one was Gemma 4 31B IT on 2026-04-02, at 47.3.
Reading these numbers
The 11 scores on this page come from 1 source: Epoch AI Benchmarking Hub (11). No official leaderboard supplied any of them, so treat them as third-party measurements.
Models ranked nearby
- #429 yi-coder:9b-chat-q4_0 (47.5)
- #430 Llama 3.1-70B (47.5)
- #431 gemini-2.0-flash-exp (47.5)
- #433 GPT-5.1 (Thinking, None) (47.5)
- #434 Claude 3.7 (47.4)
- #435 Llama 4 Scout (47.4)
More from Google DeepMind
- Gemini 3.8 Flash (high) (#26, 54.0)
- Gemini 3.7 Flash (high) (#27, 53.9)
- Gemini 3 Pro Preview (#31, 53.7)
- Gemini 3 Deep Think (#60, 52.6)
- AI co-mathematician (#69, 52.3)
- Gemini 3 Flash (#120, 51.5)
Compare it on the live leaderboard → · How the aggregate score works