Gemma 4 26B A4B gemma-4-26b-a4b-it
google · gemma efficient
#47 overall
- params
- 26.5B (A4B)
- arch
- moe
- context
- 256k
- license
- apache-2.0 open weights
- released
- Mar 2026
- downloads/30d
- 8.2M
# architecture
- layers
- 30
- d_model
- 2816
- heads
- 16q / 8kv
- head dim
- 256
- experts
- top-0 of 128
- vocab
- 262144
- family
- gemma4_text
signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.
# why ranked
Overall: #47 — score 55.3 ● high — all signals present
| signal | weight | input (0–100) |
|---|---|---|
| aa_intelligence | 0.70 | 56.1 |
| bench_composite | 0.30 | 53.4 |
benchmark panel evidence:
| complete panel | score (0–100) |
|---|---|
| aa-coding-v1 | 57.2 |
| aime-2026-v1 | 49.6 |
Coding: provisional — score 47.1 ● low — a single signal; treat with caution
provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.
| signal | weight | input (0–100) |
|---|---|---|
| aa_coding | 0.30 | 47.1 |
| aider_polyglot | 0.30 | — |
| swe_bench_verified | 0.40 | — |
missing signals are dropped and the remaining weights renormalized — never imputed.
# trend
methodology changed during this history window; score movement across that boundary is not model movement. See methodology v6.
# benchmarks
| benchmark | score | source | date |
|---|---|---|---|
| AA Coding Index | 39.3 | Artificial Analysis | — |
| AA Intelligence Index | 16.7 | Artificial Analysis | — |
| AIME 2026 | 0.9 / 1 | LLM Stats | — |
| Big-bench-extra-hard | 0.6 / 1 | LLM Stats | — |
# where to run
| provider | quant | ctx | $/M in | $/M out | $/M cache | price src | tps | uptime | ✓ |
|---|---|---|---|---|---|---|---|---|---|
| Featherless | — | — | — | — | via hfrouter | — | — | ||
| Darkbloom | unknown | 128k | $0.04 | $0.22 | $0.02 | via openrouter | — | 100.0% | |
| NextBit | bf16 | 256k | $0.07 | $0.23 | $0.04 | via openrouter | — | 100.0% | |
| Makora | unknown | 250k | $0.07 | $0.24 | $0.05 | via openrouter | — | 100.0% | |
| DekaLLM | bf16 | 256k | $0.06 | $0.33 | $0.04 | via openrouter | — | 100.0% | |
| DeepInfra | fp8 | 256k | $0.07 | $0.34 | — | deepinfra | — | 99.9% | ✓ |
| Io Net | bf16 | 262k | $0.09 | $0.31 | $0.04 | via openrouter | — | 100.0% | |
| Wandb | 262k | $0.10 | $0.30 | — | via litellm | — | — | ||
| Cloudflare | unknown | 250k | $0.10 | $0.30 | $0.05 | via openrouter | — | 100.0% | |
| CoreWeave | bf16 | 256k | $0.10 | $0.30 | $0.05 | via openrouter | — | 99.7% | |
| Amazon Bedrock | — | $0.13 | $0.40 | — | bedrock | — | — | ✓ | |
| Novita | bf16 | 256k | $0.13 | $0.40 | — | novita | — | 100.0% | ✓ |
| Parasail | bf16 | 256k | $0.13 | $0.40 | $0.05 | via openrouter | — | 100.0% | |
| Venice | bf16 | 250k | $0.13 | $0.40 | — | venice | — | 100.0% | ✓ |
| Aihubmix | 256k | $0.14 | $0.40 | — | via litellm | — | — | ||
| SiliconFlow | fp8 | 256k | $0.14 | $0.40 | $0.05 | openrouter | — | 96.2% | ✓ |
| unknown | 256k | $0.15 | $0.60 | — | via openrouter | — | 99.5% | ||
| Google Vertex | — | $0.15 | $0.60 | — | via litellm | — | — | ||
| Scaleway | — | $0.28 | $0.57 | — | via hfrouter | 172 | — |
sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price
source aliases
- aa
gemma-4-26b-a4b- arena
gemma-4-26b-a4b- bedrock
google.gemma-4-26b-a4b- hf
google/gemma-4-26B-A4B- openrouter
google/gemma-4-26b-a4b-it