Gemma 4 26B A4B
- Input
- $0.10/M tok
- Output
- $0.40/M tok
- Blended
- $0.175/M tok · 3:1
- Speed
- 75tok/s
- Context
- 256Ktokens
- Cost per task
- ~$0.17/taskest.
Long-tail import from Artificial Analysis (2026-09-24); AA variant: Gemma 4 26B A4B (Reasoning). Price = AA median host price.
Rankings
Coding
Verified#90 / 137rank by quality36.8±10.1quality / 1001/7 benchmarks measuredWriting & chat
Verified#47 / 162rank by quality62.5±4.5quality / 1003/5 benchmarks measuredPrice frontierVision
Verified#55 / 98rank by quality51.3±8.2quality / 1001/6 benchmarks measuredHard reasoning
Verified#99 / 218rank by quality48.2±5.9quality / 1002/4 benchmarks measuredAgents
Verified#105 / 187rank by quality27.8±5.9quality / 1003/7 benchmarks measured
Self-hostingEstimated
How we estimate →What it would cost to run these open weights yourself on rented GPUs, compared with the API price above.
- Hardware
- 1× B200 180GB
- BF16 weights
- Throughput
- ~10,963 tok/s
- many requests batched
- Price
- $0.062–0.56 /M tok
- busy → light use
- On your own machine
- RTX 5090 32GB (int4, 300+ tok/s single-stream)
- single consumer GPU or Mac
Assumptions (5)
- BF16 weights (25.2B params, 3.8B active per token) + 25% KV-cache headroom ≈ 63 GB
- 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
- ~10,963 output tok/s aggregate at batch 256 (bandwidth-bound); MoE compute scales with active params
- Blended 3:1 input:output; prefill ~139,885 tok/s
- Low = 75% utilization at the low GPU price; high = 20% at the high price
Benchmark scores
~ italic, dashed = no published score yet, estimated from related benchmarks.
Coding
SWE-bench VerifiedEstimated
~#17 of 21 (estimated)~70.1%
- #1Claude Opus 596.0%
- #2Claude Sonnet 585.2%
- #16Nemotron 3 Ultra70.7%
- ~#17Gemma 4 26B A4BEstimated70.1%
- #17Intern-S2-Preview (35B-A3B)64.0%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#23 of 24 (estimated)~49.0%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #22Muse Glimmer51.2%
- ~#23Gemma 4 26B A4BEstimated49.0%
- #23Nex-N2.5-Mini43.8%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#41 of 46 (estimated)~40.7%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #40Qwen3.8 27B42.2%
- ~#41Gemma 4 26B A4BEstimated40.7%
- #41Laguna S 2.140.4%
No published score — estimated from related benchmarks
Terminal-Bench 4.0Estimated
~#40 of 102 (estimated)~9.9%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #39Qwen3.8 2.4T A95B11.1%
- ~#40Gemma 4 26B A4BEstimated9.9%
- #40Muse Spark Open8.1%
No published score — estimated from related benchmarks
LiveCodeBenchEstimated
~#15 of 28 (estimated)~65.4%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #14Maple-Preview75.1%
- ~#15Gemma 4 26B A4BEstimated65.4%
- #15Mistral Small 464.0%
No published score — estimated from related benchmarks
SciCodeVerified
#80 of 11840.0%
- #1Claude Fable 5.267.7%
- #2Claude Opus 5.566.9%
- #79Mistral Medium 3.540.2%
- #80Gemma 4 26B A4B40.0%
- #80K2 Horizon MoVA 36B A4B40.0%
artificialanalysis.ai · 2026-09-24
AA Coding Agent IndexEstimated
~#24 of 26 (estimated)~31.2
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #23Gemini 3.8 Flash33.3
- ~#24Gemma 4 26B A4BEstimated31.2
- #24GPT-6 Luna31.1
No published score — estimated from related benchmarks
Writing & chat
IFBenchVerified
#30 of 13972.4%
- #1MiniMax-M382.9%
- #2Qwen3.8 2.4T A95B82.8%
- #29LFM3-1.7B73.0%
- #30Gemma 4 26B A4B72.4%
- #31Nemotron 3 Super 120B A12B71.5%
artificialanalysis.ai · 2026-09-24
LMArena Text, Non-English (Elo)Few votesVerified
#34 of 561423 Elo
- #1Gemini 4 ArgonFew votes1512
- #2Claude Opus 5.51497
- #33MiniMax-M31427
- #34Gemma 4 26B A4BFew votes1423
- #35Muse GlimmerFew votes1418
lmarena.ai · 2026-09-13
EQ-Bench Creative Writing v3 (Elo)Estimated
~#13 of 26 (estimated)~1805 Elo
- #1GPT-6 Astra2173
- #2Claude Fable 5.12162
- #12Qwen3.8 2.4T A95B1843
- ~#13Gemma 4 26B A4BEstimated1805
- #13Muse Glimmer1798
No published score — estimated from related benchmarks
Vision
CharXiv ReasoningEstimated
~#11 of 12 (estimated)~76.9%
- #1Gemini 3.8 Flash86.2%
- #2Kimi K384.8%
- #10Claude Sonnet 577.0%
- ~#11Gemma 4 26B A4BEstimated76.9%
- #11Gemini 3.5 Flash-Lite74.5%
No published score — estimated from related benchmarks
OmniDocBench v1.5Estimated
~#5 of 8 (estimated)~83.0%
- #1MiniMax-M391.6%
- #2Kimi K391.1%
- #4Gemini 3.1 Pro85.3%
- ~#5Gemma 4 26B A4BEstimated83.0%
- #5Claude Haiku 4.579.6%
No published score — estimated from related benchmarks
GDP.pdf (AA)Estimated
~#33 of 48 (estimated)~11.5%
- #1Muse Wonder Full32.6%
- #2GPT-6 Astra31.0%
- #31Qwen3.7 Plus12.2%
- ~#33Gemma 4 26B A4BEstimated11.5%
- #33Muse Glimmer10.0%
No published score — estimated from related benchmarks
ChartographyEstimated
~#20 of 25 (estimated)~15.1%
- #1Claude Fable 5.273.4%
- #2GPT-6 Astra71.0%
- #19GLM-5.3 Flash16.3%
- ~#20Gemma 4 26B A4BEstimated15.1%
- #20Grok 4.714.7%
No published score — estimated from related benchmarks
Hard reasoning
GPQA DiamondVerified
#81 of 19279.2%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #80EXAONE 4.5 33B79.4%
- #81Gemma 4 26B A4B79.2%
- #82Nova 2.0 Pro Preview78.5%
artificialanalysis.ai · 2026-09-24
Humanity's Last ExamVerified
#83 of 19819.3%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #82gpt-oss-120b19.6%
- #83Gemma 4 26B A4B19.3%
- #84Gemini 3.5 Flash-Lite18.8%
artificialanalysis.ai · 2026-09-24
AIME (latest)Estimated
~#22 of 34 (estimated)~80.0%
- #1GLM-5.3 NVFP499.5%
- #2Beam 1.0 501B A23B97.7%
- #21Magistral Small 1.280.3%
- ~#22Gemma 4 26B A4BEstimated80.0%
- #22Phi-4-reasoning-plus78.0%
No published score — estimated from related benchmarks
AA Intelligence IndexEstimated
~#58 of 103 (estimated)~21.8
- #1Gemini 4 Carbon59.4
- #2Claude Opus 5.557.6
- #57Gemini 3.5 Flash-Lite22.2
- ~#58Gemma 4 26B A4BEstimated21.8
- #58MiMo-V2.521.8
No published score — estimated from related benchmarks
Agents
τ²-bench (Telecom)Verified
#54 of 11943.6%
- #1Step 3.7 Flash98.5%
- #2DeepSeek-V4-Pro-0813-nvfp4-DSpark98.3%
- #53Nemotron 3 Nano Omni 30B A3B Reasoning45.3%
- #54Gemma 4 26B A4B43.6%
- #55Mistral Small 441.2%
artificialanalysis.ai · 2026-09-24
OSWorld-VerifiedEstimated
~#12 of 15 (estimated)~62.3%
- #1Kimi K384.8%
- #2Qwen3.8 27B84.3%
- #11Muse Glimmer65.9%
- ~#12Gemma 4 26B A4BEstimated62.3%
- #12Nemotron 3.5 Super VL NVFP452.3%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~last of 5 (estimated)below measured range (<47.5%)
No published score — estimated from related benchmarks
BrowseCompEstimated
~#19 of 22 (estimated)~57.8%
- #1Atria Dawn Preview92.5%
- #2GPT-6 Astra91.5%
- #18Nemotron 3.5 Bolt 74B-A9B70.0%
- ~#19Gemma 4 26B A4BEstimated57.8%
- #19Mistral Medium 3.548.6%
No published score — estimated from related benchmarks
GDPval-AA (Elo)Verified
#65 of 91553 Elo
- #1Claude Opus 5.51866
- #2Muse Wonder Full1857
- #64Granite 4.2 30B564
- #65Gemma 4 26B A4B553
- #66Devstral 2531
artificialanalysis.ai · 2026-09-24
AutomationBench-AAEstimated
~#45 of 85 (estimated)~16.1%
- #1Claude Fable 5.276.4%
- #2Muse Wonder Full71.3%
- #44Qwen3.7 Plus17.4%
- ~#45Gemma 4 26B A4BEstimated16.1%
- #45Ling-3.0-flash-VL15.7%
No published score — estimated from related benchmarks