We shape the world's best technologies to help those learning the least.

The platform for AI for education in LMICs.

Scalable quality assurance.
Real-world innovation.

AI model performance.

Benchmark Leaderboard
# ↕	Model ↕	Pedagogy ▼	Cost ($/1M Tokens) ↕	Size (Bn Params) ↕	Open
1	Gemini-2.5 Pro	89%	$1.25	—	No
2	Grok-4	88%	$3	—	No
3	o3	88%	$2	—	No
4	GPT-5	88%	$1.25	—	No
5	Claude Opus 4 Thinking	87%	$15	—	No
6	Claude Sonnet 4 Thinking	87%	$3	—	No
7	Deepseek R1 (May '25)	87%	$3	671	Yes
8	Claude Opus 4	86%	$15	—	No
9	GLM-4.5	86%	$0.55	355	Yes
10	Grok-4 Fast	86%	$0.2	—	No
11	Gemini-2.5 Flash	86%	$0.3	—	No
12	Qwen-3 235B A22B Thinking	86%	$0.22	235	Yes
13	Claude Sonnet 4	85%	$3	—	No
14	GPT-5 Mini	83%	$0.25	—	No
15	Qwen-3 32B	82%	$0.1	32	Yes
16	Qwen-3 235B A22B	82%	$0.15	235	Yes
17	GPT-4.1	82%	$2	—	No
18	Kimi K2	82%	$0.6	1000	Yes
19	o4-Mini	82%	$1.1	—	No
20	Grok-3	82%	$3	—	No
21	Llama-4 Maverick	82%	$0.2	402	Yes
22	Mistral Medium 3	80%	$0.4	—	No
23	Gemini-2.0 Flash	80%	$0.1	—	No
24	Qwen-3 30B A3B	80%	$0.1	30	Yes
25	o3-Mini	79%	$1.1	—	No
26	Llama-3.1 405B	79%	$3	405	Yes
27	Qwen-3 14B	79%	$0.07	14	Yes
28	GPT-4o	78%	$2.5	—	No
29	Gemini-2.0 Flash-Lite	78%	$0.075	—	No
30	Claude-3.7 Sonnet	77%	$3	—	No
31	Llama-3.2 90B	76%	$0.9	90	Yes
32	Qwen-2.5 72B	75%	$0.9	72	Yes
33	Nova Pro	75%	$0.8	—	No
34	Mistral Large November	75%	$2	123	Yes
35	GPT-4.1 Mini	75%	$0.4	—	No
36	Magistral Medium	75%	$2	—	No
37	Llama-3.1 70B	75%	$0.9	70	Yes
38	GLM-4.5 Air	75%	$0.22	106	Yes
39	Llama-4 Scout	75%	$0.15	109	Yes
40	Qwen-3 8B	74%	$0.035	8	Yes
41	Gemma-3 27B	73%	$0.3	27	Yes
42	GPT-5 Nano	72%	$0.05	—	No
43	Mistral Small 3	70%	$0.05	24	Yes
44	Qwen-3 4B	70%	—	4	Yes
45	GPT-4o Mini	69%	$0.15	—	No
46	Gemma-3 12B	68%	$0.05	12	Yes
47	Qwen-2.5 7B	67%	$0.04	7	Yes
48	GPT-4.1 Nano	64%	$0.1	—	No
49	Claude-3.5 Haiku	63%	$1	—	No
50	LFM-7B	62%	$0.01	7	No
51	Jamba 1.6 Mini	61%	$0.2	52	Yes
52	Qwen-3 1.7B	58%	—	1.7	Yes
53	Mistral Nemo	57%	$0.01	12	Yes
54	Gemma-3 4B	52%	$0.02	4	Yes
55	GPT-3.5 Turbo	52%	$0.5	—	No
56	Command R7B	46%	$0.0375	7	Yes
57	Llama-3.2 1B	28%	$0.005	1	Yes
58	SmolLM2-135M	21%	—	0.135	Yes

QA framework from development to scale background

A QA framework from development to scale.

From solution development through implementation, understanding the user experience to measuring the impact. Our framework looks at how to assure quality throughout the lifecycle of a product.

Research & Development

Small grants to support high potential solutions.

“The EGRA-AI project uses a reading assessment tool that uses automatic speech recognition (ASR) to evaluate early grade reading.”

María José Ogando Portela

Director, Fab AI

See case study

Education is, at its core, human. Teachers and parents inspire learning; people shape the system.

At Fab AI, we use technology not to replace, but to supercharge, the human connections that make education work.

We shape the world's best technologies to help those learning the least.

Scalable quality assurance.
Real-world innovation.

AI model performance.

Leaderboard

Plot charts

About

A QA framework from development to scale.

Research & Development

Education is, at its core, human. Teachers and parents inspire learning; people shape the system.

EdTech Quality

Implementation

Sign up for AI-for-Education.org news updates

We shape the world's best technologies to help those learning the least.

Scalable quality assurance.Real-world innovation.

AI model performance.

Leaderboard

Plot charts

About

A QA framework from development to scale.

Research & Development

Education is, at its core, human. Teachers and parents inspire learning; people shape the system.

Scalable quality assurance.
Real-world innovation.