Model
Model explorer

Flan-T5 XXL

OPEN
Google · Flan-T5 family · released Oct 20, 2022

11B instruction-tuned T5; popularized the Flan instruction-tuning recipe on fully open weights.

ReasoningCodingVisionFunction callingTool useAgentic
108.0
Elo · rank #381
Parameters
11B
Active params
11B (dense)
Context
1K tokens
Architecture
Dense encoder-decoder Transformer (11B, T5 + Flan instruction tuning)
License
Apache 2.0
Languages
—
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
BIG-Bench HardReasoning45.3%#103
best: ERNIE 4.5 300B-A47B · 94.3%
GPQA DiamondReasoning27.0%#269
best: GPT-5.6 · 94.6%
IFEvalReasoning22.0%#151
best: Gemma 4 26B A4B · 98.5%
MGSMMath4.9%#52
best: OpenAI o4-mini · 93.7%
MMLU-ProKnowledge23.4%#169
best: Claude Fable 5 · 91.5%
MMLUKnowledge55.1%#221
best: OpenAI o3 · 92.9%
Run it locally
VRAM @ Q4
7 GB
VRAM @ FP16
22 GB
Fits on (Q4)
RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
T5/encoder-decoder support in llama.cpp is limited/experimental; no widely measured RTX 4090 llama.cpp throughput figures found.
Quantizations
GGUF Q4 · GGUF Q8
Flan-T5 family
Elo progression across releases
Flan-T5 XXLOct 2022108.0
Resources
HFgoogle/flan-t5-xxlGHgoogle-research/FLANDOCSarxiv.org/abs/2210.11416
API price weights · each benchmark row carries its own source badge (see methodology)