Model
Model explorer

Falcon3-7B

OPEN
TII · Falcon 3 family · released Dec 17, 2024

From-scratch 14T-token anchor of Falcon 3 with 32K context; strong math/code for its size.

ReasoningCodingVisionFunction callingTool useAgentic
419.6
Elo · rank #327
Parameters
7B
Active params
7B (dense)
Context
32K tokens
Architecture
Dense decoder-only transformer (28 layers, GQA 12/4 heads, SwiGLU)
License
TII Falcon-LLM License 2.0
Languages
4+
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
AlpacaEvalHuman preference26.1%#33
best: Tülu 2+DPO 70B · 95.1%
ARC-ChallengeReasoning62.6%#77
best: Llama 3.1 405B · 96.9%
BIG-Bench HardReasoning52.4%#95
best: ERNIE 4.5 300B-A47B · 94.3%
BFCLAgents89.5%#3
best: SmolLM3 3B (No Thinking) · 92.3%
GPQA DiamondReasoning31.9%#254
best: GPT-5.6 · 94.6%
GSM8KMath81.4%#75
best: Llama 3.1 405B · 96.8%
IFEvalReasoning76.5%#103
best: Gemma 4 26B A4B · 98.5%
MATH-500Math29.4%#167
best: GPT-5 · 99.4%
MMLU-ProKnowledge40.7%#148
best: Claude Fable 5 · 91.5%
MMLUKnowledge70.5%#142
best: OpenAI o3 · 92.9%
MT-BenchHuman preference8.4#31
best: Hunyuan-Large (A52B) · 9.4
OpenBookQAReasoning45.8%#35
best: Claude 1 · 90.8%
PIQAReasoning78.8%#50
best: GPT-4o mini · 93.1%
WinoGrandeReasoning70.4%#99
best: PaLM 2 · 90.9%
Run it locally
VRAM @ Q4
4.6 GB
VRAM @ FP16
16.52 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
Throughput data unavailable.
Quantizations
GGUF · AWQ · GPTQ · MLX
Falcon 3 family
Elo progression across releases
Falcon3-10BDec 2024421.4Falcon3-7BDec 2024419.6
Resources
HFtiiuae/Falcon3-7B-InstructDOCSfalconllm.tii.ae/falcon3APIAvailable via TII API
API price weights · each benchmark row carries its own source badge (see methodology)