Gemma 4 26B A4B
GoogleReleased Apr 2, 2026Open weights · Apache 2.0gemma-4-26b-a4b
- Intelligence (Epoch)
- 142 · 76th of 111
- Reasoning
- yes
- Knowledge cutoff
- January 2025
- Context
- 262K tokens
- Max output
- 128k tokens
- Input
- text, image, video, audio
- API features
- tools, structured outputs, JSON mode, seed, logprobs
- Google's API
- yes, as gemma-4-26b-a4b-it-maas
Read Oct 23, 06:00 UTC from the hosts' published offers and OpenRouter, spec sheet from Google · cutoff: ai.google.dev
Free right now
- Free OpenRouter, served by Google AI Studio · no end date · 20 req/min, 50 req/day, 1,000 req/day with $10 of credits
Checked Oct 23 · every free model on Deals
The cheapest host, Reka, serves a weights precision we can't read (host terms). Cheapest at the maker's precision and context: NextBit, $0.33 blended. Weights and context per host
Intelligence for the price
4 models score higher for the same price or less; it ranks 5 of the 15 models at its price or below.
Epoch Capabilities Index against the lowest input + output price per M, log scale · prices Oct 23. Score: Epoch AI, CC BY 4.0, 142 with an interval of 139 to 143.
Against its rivals
The best of its price range and the top score
- Gemma 4 26B A4Bthis model$0.26142
- Claude Opus 5.5top score$24.00169
- DeepSeek V4.1 Flashsame price range$0.47156
- DeepSeek V4 Flash 0731same price range$0.21155
Same level for 2.2× less: gpt-oss-20b at AkashML, $0.12 per M input + output · their Epoch intervals overlap (138 against 142).
Bars: Epoch Capabilities Index (Epoch AI, CC BY 4.0), drawn from 100, below the lowest score in the catalogue · price: lowest input + output per M.
Speed by provider
Output speed of Gemma 4 26B A4B at each host, fastest first
- DekaLLM64 tok/s0.56 s to first token
- NextBit59 tok/s0.34 s to first token
- DeepInfra58 tok/s0.44 s to first token
- Io Net54 tok/s0.69 s to first token
- Cloudflare48 tok/s0.60 s to first token
- Parasail39 tok/s0.86 s to first token
- CoreWeave37 tok/s0.38 s to first token
- SiliconFlow31 tok/s1.97 s to first token
- Venice30 tok/s1.11 s to first token
- Google30 tok/s1.50 s to first token
Speed: OpenRouter, median of the last 30 min, read Oct 23, 06:00 UTC · a snapshot of real traffic before the daily read, not a daily average · bars from zero.
Run it yourself
open weights, Apache 2.0 · fits 6 of the 6 workstations in our guideSelf-host guide ›- Parameters
- 26B
- 4B active per token (MoE)
- Memory needed
- 16 GB
- in INT4 · 31 GB in INT8
- Smallest cloud setup
- RTX 3090
- 24 GB · $0.17/h, about $124 a month
- Worth it above
- 477M tokens
- a month, against the cheapest API at $0.26 per M
| Quantization | Memory | Smallest cloud GPU | Quality |
|---|---|---|---|
| INT4 (Q4) | 16 GB | RTX 3090 · $0.17/h | smallest footprint, minor quality loss |
| INT8 (Q8) | 31 GB | A100 · $1.19/h | near-lossless |
| FP16 (full) | 62 GB | A100 · $1.19/h | reference quality |
At home: in INT4 it fits a RTX 5090 (32 GB) + host PC (plus a host PC), about $3,500 once, or $0.1332/h over 3 years, electricity aside; it pays for itself after about 13B tokens against the cheapest API.
Memory ≈ parameters × bytes per parameter × 1.2 · cloud price: on-demand rate, running 24/7 · builds: GGUF, AWQ, MLX, GPTQ, Ollama · break-even calculator
| Provider | Input | Output | Cache ↑ | tok/s | First token | Uptime 24h | Checked |
|---|---|---|---|---|---|---|---|
| Rekaprecision not read: host terms · 256K ctx · 128K out | $0.06 | $0.20 | $0.035 | — | — | 99.8% | Sep 30 |
| DekaLLMbf16 · 256K ctx | $0.06 | $0.33 | $0.04 | 64 | 0.56 s | 99.9% | Oct 23 |
| Io Netbf16 · 262K ctx | $0.09 | $0.31 | $0.041 | 54 | 0.69 s | 100% | Oct 23 |
| NextBitbf16 · 256K ctx · 16K out−15%reprices often | $0.0765 | $0.26 | $0.0425 | 59 | 0.34 s | 100% | Oct 23 |
| Cloudflareprecision not stated · 250K ctx | $0.10 | $0.30 | $0.05 | 48 | 0.60 s | 100% | Oct 23 |
| CoreWeavebf16 · 256K ctx | $0.10 | $0.30 | $0.05 | 37 | 0.38 s | 100% | Oct 23 |
| Venicebf16 · 250K ctx · 8K out | $0.13 | $0.40 | $0.05 | 30 | 1.11 s | 99.9% | Oct 23 |
| Parasailbf16 · 256K ctx | $0.13 | $0.40 | $0.05 | 39 | 0.86 s | 99.8% | Oct 23 |
| SiliconFlowfp8 (bf16) · 256K ctx | $0.14 | $0.40 | $0.05 | 31 | 1.97 s | 97.3% | Oct 23 |
| DeepInfrafp8 (bf16) · 256K ctx · 16K out | $0.07 | $0.34 | — | 58 | 0.44 s | 99.7% | Oct 23 |
| Novita AIbf16 · 256K ctx · 128K out | $0.13 | $0.40 | — | — | — | — | Oct 23 |
| Googleprecision not stated · 256K ctx · 125K out | $0.15 | $0.60 | — | 30 | 1.50 s | 94.4% | Oct 23 |
10 of 12 providers · prices per million tokens, read Oct 23, 06:00 UTC · speed: OpenRouter, median of the last 30 min, read Oct 23, 06:00 UTC · uptime: OpenRouter, last 24 h · −N%: a discount off the host's list price · under each host: weights precision, context and max output as served, the maker's value in brackets when the host's differs
| Provider | Weights | Context | Max output | Tools | Structured outputs | Reasoning |
|---|---|---|---|---|---|---|
| Reka | not read: host terms | 256K | 128K | — | — | — |
| NextBit | bf16 | 256K | 16K | yes | yes | yes |
| DekaLLM | bf16 | 256K | not checked yet | yes | no | yes |
| Cloudflare | not stated by the host | 250K | not stated by the host | yes | no | yes |
| CoreWeave | bf16 | 256K | not checked yet | no | yes | yes |
| Io Net | bf16 | 262K | not read: host terms | yes | no | yes |
| DeepInfra | fp8 | 256K | 16K | yes | yes | yes |
| Venice | bf16 | 250K | 8K | yes | yes | yes |
| Parasail | bf16 | 256K | not read: host terms | no | yes | yes |
| Novita AI | bf16 | 256K | 128K | yes | no | yes |
| SiliconFlow | fp8 | 256K | not read: host terms | no | yes | yes |
| not stated by the host | 256K | 125K | yes | yes | yes |
Weights, context and max output: read at each host's own catalogue or docs where its terms allow, otherwise as the host declares to OpenRouter; hover a cell for its source and date · tools, structured outputs, reasoning: as declared to OpenRouter · not verified by us
Gemma 4 26B A4B spec sheet, field by field, with sources
Gemma 4 26B A4B spec sheetmaker-declared
What Google declares about the model itself — read from its own weights repository and documentation, not from any host. What each endpoint actually serves is in the pricing above: the two can differ, and both are true.
- Context window
- 262,144 tokens
- Parameters
- 25,805,936,206≈ 25.8B
- Active experts
- none declared in the weights
- Weight precision
- BF16
- License
- Apache 2.0 — open weights, permissivecurated · read 2026-09-07Scope: the licence on the published weights, and nothing else. It is not the maker's acceptable-use policy, which is a separate document, and it is not the contract you sign with whichever host you call — hosts set their own terms, which this index does not read.
- Knowledge cutoff
- January 2025maker's docs
- Modalities
- text · image · video · audio
- Max output (official)
- to verify
- Release date
- to verify≈ April 2, 2026 third-party catalogue — not the maker's figure
- Maker lifecycle status
- to verify
Also declared in the weights
- Architecture class
- Gemma4ForConditionalGeneration
- Layers
- 30
- Vocabulary
- 262,144 tokens
- Weights first published
- March 11, 2026
- Weights last updated
- July 20, 2026
- Paper
- arXiv:2607.02770
google/gemma-4-26B-A4B-it · config.json @ 4d7ae49 · read 2026-10-08 · 6 of 10 fields published by the maker
| Date | Provider | Input | Output | Change |
|---|---|---|---|---|
| Oct 23 | NextBit | $0.09 → $0.0765 | $0.30 → $0.26 | −15% |
| Oct 22 | NextBit | $0.0765 → $0.09 | $0.26 → $0.30 | +18% |
| Oct 20 | NextBit | $0.0675 → $0.0765 | $0.23 → $0.26 | +13% |
| Oct 17 | NextBit | $0.0765 → $0.0675 | $0.26 → $0.23 | −12% |
| Oct 16 | Io Net | $0.11 → $0.09 | $0.32 → $0.31 | −7% |
| Oct 15 | Io Net | $0.12 → $0.11 | $0.35 → $0.32 | −9% |
| Oct 13 | Io Net | $0.0483 → $0.12 | $0.21 → $0.35 | +83% |
| Oct 13 | NextBit | $0.0675 → $0.0765 | $0.23 → $0.26 | +13% |
| Oct 10 | NextBit | $0.0765 → $0.0675 | $0.26 → $0.23 | −12% |
| Oct 9 | Io Net | $0.0517 → $0.0483 | $0.23 → $0.21 | −7% |
| Oct 8 | Io Net | $0.0587 → $0.0517 | $0.26 → $0.23 | −12% |
| Oct 7 | Io Net | $0.0621 → $0.0587 | $0.27 → $0.26 | −6% |
| Oct 6 | Io Net | $0.15 → $0.0621 | $0.50 → $0.27 | −49% |
| Oct 6 | NextBit | $0.0675 → $0.0765 | $0.23 → $0.26 | +13% |
| Oct 3 | NextBit | $0.0765 → $0.0675 | $0.26 → $0.23 | −12% |
| Sep 29 | NextBit | $0.0675 → $0.0765 | $0.23 → $0.26 | +13% |
| Sep 26 | NextBit | $0.09 → $0.0675 | $0.30 → $0.23 | −25% |
| Sep 13 | NextBit | $0.10 → $0.09 | $0.40 → $0.30 | −22% |
| Sep 10 | SiliconFlow | $0.12 → $0.14 | $0.40 → $0.40 | +4% |
Every published change of input or output price, newest first · change on input + output · full series: pricing-history.csv.
| Provider | Chatbot 1,000 in / 500 out | RAG / long context 8,000 in / 500 out | Batch summarize 4,000 in / 1,000 out |
|---|---|---|---|
| Reka | $0.16 | $0.58 | $0.44 |
| NextBit | $0.20 | $0.74 | $0.56 |
| DekaLLM | $0.23 | $0.64 | $0.57 |
| Cloudflare | $0.25 | $0.95 | $0.70 |
| CoreWeave | $0.25 | $0.95 | $0.70 |
| Io Net | $0.24 | $0.88 | $0.67 |
| DeepInfra | $0.24 | $0.73 | $0.62 |
| Venice | $0.33 | $1.24 | $0.92 |
| Parasail | $0.33 | $1.24 | $0.92 |
| Novita AI | $0.33 | $1.24 | $0.92 |
| SiliconFlow | $0.34 | $1.32 | $0.96 |
| $0.45 | $1.50 | $1.20 |
What each host bills for 1,000 requests of a representative job
Your own workload
Other Google models · all its models, hosts and history ›
- Gemma 4 31B14 providers · from $0.38
- Gemma 3 27B5 providers · from $0.24
- Gemma 3 12B3 providers · from $0.15
- Gemma 3 4B2 providers · from $0.09
- Gemini 2.5 Flash1 provider · from $2.80
- Gemini 2.5 Pro1 provider · from $11.25