GPT-4.1 nano

+ contribute

GPT-4.1 nano is a closed-weights language model from OpenAI, released April 2025. Licence: proprietary.

Catalogue identity notes

Comparison model from the Epoch AI internal-evaluation CSVs (Epoch release date 2025-04-14). Dated API snapshot gpt-4.1-nano-2025-04-14 (the only snapshot); sibling of gpt-4-1 / gpt-4-1-mini. OpenAI's deprecations page schedules its shutdown for October 23, 2026 (announced 2026-04-22).

52.8±0.7#302 of 719 LLMsfrom 35 benchmark results across 17 sources

$ how it compares

The LLMs ranked either side of GPT-4.1 nano, out of 828 with a score, then more from OpenAI.

#	model	score	where	RAM	released
360	DeepSeek-Coder-V2 Instruct (0724) DeepSeek	53±1	local	~151GB	2024-09
361	Olmo 3 7B Think Ai2	53±1	local	~5.8GB	2025-11
362	GPT-4.1 nano OpenAI	53±1	api $0.17/M		2025-04
363	Gemini 1.5 Flash (002) Google	53±1	comparison only		2024-09
364	Gemma 3 27B Google	53±1	api $0.10/M +local	~18GB	2025-03
	more from OpenAI
1	GPT-6 Astra OpenAI	87±1	api $20.00/M		2026-09
3	GPT-6.1 Sol OpenAI	85±1	api $4.00/M		2026-09
7	GPT-6 Sol OpenAI	83±1	api $4.00/M		2026-09

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #240 of 300 · 8.0% of its score · 7 tests

Beats DeepSeek R1 0528 Qwen3 8B, Sky-T1 32B Preview, Magistral Small 1.0 (2506) here, though they rank higher overall. Trails Granite 4.0 1B, SmolLM3 3B, Phi-3.5-mini-instruct here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
BFCL V4 (December 2025 archive)	33.1% ✎	▁▁▁▂██▅▅▆▂▂▂ beats 37/77	hard · avg 32%	▇ 2.1%	at its level
τ²-bench Telecom lab	12.1% ✎	▃·▅▅▃▃··█▃▃▅ beats 0/13	hard · avg 16%	▇ 1.7%	sharp · noisy
Finance Agent Benchmark v1 (537 questions, class-balanced accuracy)	2.4% ✎	█·▃▃·█▅▅▅▅▃▅ beats 0/18	very hard · avg 5.8%	▇ 1.4%	sharp · too hard for it
BFCL v3	53.0% ✎	▁▁▁▁▁▃▅█▅▄▁▁ beats 50/114	medium · avg 52%	▇ 1.0%	blunt · at its level
Galileo Agent Leaderboard v2 (action completion)	38.0% ✎	▅▃·▃▃▅▅█▃█▅▅ beats 7/19	hard · avg 27%	▇ 0.8%	at its level
Galileo Agent Leaderboard v1 (tool selection quality)	80.3% ✎	▃▃▃▃▆▅▆▅█▂▅▂ beats 9/31	easy · avg 79%	▇ 0.8%	blunt
KlinikeBench strict pass@1 (333 interactive clinical encounters, Terminus-2)	0.9% ✎	▁▃·▁▁▁▁▅▃█▅█ beats 1/27	very hard · avg 2.4%	▇ 0.3%	too hard for it

coding▁▁▁▃▆██▇▆▄▃▁ #254 of 458 * · 8.0% of its score · 2 tests

Beats Qwen3-VL 32B Instruct, Solar Open 100B (102B-A12B), GLM-4.7-Flash here, though they rank higher overall. Trails DeepSeek LLM 67B Chat, QwQ-32B-Preview, InternLM2.5 20B Chat here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Aider polyglot (225 exercises, two attempts)	8.9% ✎	▂·▃▃▅▂▃█▆▄▂▂ beats 2/49	very hard · avg 8.5%	▇▇▇ 6.3%	sharp
HumanEval (simple-evals) lab	87.0% ✎	▆█▆▆▂··▂···▂ beats 2/15	easy · avg 84%	▇ 1.7%	typical

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #199 of 393 · 10% of its score · 3 tests

Beats Granite 4.2 8B, Qwen3-VL 30B-A3B Thinking, Mistral Small 4 (119B-A6.5B) here, though they rank higher overall. Trails LFM2.5 1.2B Instruct, LFM2.5 1.2B Thinking, Gemma 3 12B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1322.3 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 157/379	avg 1268	▇▇ 4.5%	blunt
Arena-Hard v2.0 hard prompts (style control, Gemini-2.5 judge)	13.7% ✎	▂·▃▅▅·▃█▆▂▂▂ beats 3/22	very hard · avg 12%	▇▇ 4.3%	sharp
Arena-Hard v2.0 creative writing (GPT-4.1 + Gemini-2.5 judges)	10.7% ✎	▆▂▂▆▂▂▃█▂▃·▂ beats 3/22	hard · avg 19%	▇ 1.0%	sharp · at its level · noisy

math▁▂▃▅▆▅▄█▄▁▁▁ #208 of 398 · 21% of its score · 4 tests

Beats DeepSeek V3 (Dec '24), GPT-4o (2024-08-06), Command A (03-2025) here, though they rank higher overall. Trails Olmo 3 7B Instruct, DeepSeek R1 Distill Llama 8B, Phi-4-mini-reasoning 3.8B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
MATH level 5	70.0% ✎	▂▂▅▆▄█▆▃▅▂▃▄ beats 60/92	hard · avg 32%	▇▇▇▇ 11%	sharp · at its level
MATH-500 (simple-evals) lab	62.3% ✎	▃··▅▅▃···▃·█ beats 0/9	medium · avg 61%	▇▇ 5.9%	sharp · at its level
OTIS mock AIME 2024–2025	28.9% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 63/193	very hard · avg 8.5%	▇ 2.3%	sharp · noisy
MGSM (simple-evals) lab	73.0% ✎	▃····▁▅▁█▁▃▁ beats 0/16	easy · avg 82%	▇ 1.7%	blunt

vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #115 of 147 · 11% of its score · 3 tests

Beats DeepSeek V4 Flash Vision (Exp), GPT-4 Turbo (2024-04-09), Qwen2-VL 72B Instruct here, though they rank higher overall. Trails Qwen3.5 2B, Qwen3-VL 4B Instruct, Gemma 4 E4B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
CharXiv descriptive questions lab	73.9% ✎	▄▄·····▄·▄█▄ beats 0/6	easy · avg 72%	▇▇ 4.6%	typical
Arena vision (style controlled)	1088.7 ✎	▁▁▂▂▂▃▄▅▅▆██ beats 28/145	avg 1068	▇ 3.1%	blunt
CharXiv (reasoning questions) lab	40.5% ✎	▄··▂█··▂···▄ beats 1/9	hard · avg 34%	▇ 2.9%	at its level

web search▂▁▂▆▄█▃▃▁▁·▁ #64 of 80 * · 1.4% of its score · 1 test

Beats DeepSeek V4 Flash (Preview, Apr 2026), Nemotron 3 Ultra 550B-A55B, GLM-4.7 here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
Finance Agent Benchmark v1 (537 questions, class-balanced accuracy)	2.4% ✎	█·▃▃·█▅▅▅▅▃▅ beats 0/18	very hard · avg 5.8%	▇ 1.4%	sharp · too hard for it

other tests16 not in a task · 42% of its score

test	result	vs the others in its run	difficulty	counts	why
GPQA Diamond (simple-evals) lab	50.3% ✎	█▃▃█·▃▃▅▃▃▃▅ beats 6/16	hard · avg 40%	▇▇▇ 8.1%	sharp · at its level
GPQA Diamond	48.9% ✎	▁▁▁▁▂▄▆▄▅▇█▅ beats 65/208	hard · avg 40%	▇▇▇ 6.5%	sharp · at its level
CLIcK Korean (Daekeun Kim)	61.2% ✎	▆▄▄▄█▄▂·▄█·▄ beats 6/23	medium · avg 58%	▇▇▇ 6.3%	at its level
HAE-RAE 1.0 (Daekeun Kim)	68.0% ✎	▄▄▄▃▂█▂▄▂█▂▄ beats 6/22	medium · avg 63%	▇▇ 5.3%	at its level
MMLU (simple-evals, zero-shot CoT) lab	80.1% ✎	▃▂·█▄▄▂·█▂·▂ beats 0/16	easy · avg 80%	▇ 3.5%	typical
EQ-Bench 3 (emotional intelligence, judge Elo)	1134.9 ✎	▁·▂▁▄▅▅▃█▄▁▁ beats 16/62	avg 1044	▇ 2.8%	typical
HindiGen-v1 3C3H score (Inception, Claude 3.5 Sonnet judge)	56.9% ✎	▁▁▁▁·▂·█▆█▇▁ beats 12/31	medium · avg 48%	▇ 2.3%	at its level
KMMLU-HARD (Daekeun Kim, zero-shot)	24.3% ✎	█···▁▁▁▂▃·▆▂ beats 2/23	very hard · avg 29%	▇ 1.9%	sharp · too hard for it

8 more tests

test	result	vs the others in its run	difficulty	counts	why
Open Persian LLM Leaderboard v2 average (21 tasks)	60.1% ✎	▂▄▂▃█▄▃▆▃▅▂▂ beats 50/81	medium · avg 58%	▇ 1.9%	blunt · at its level
ARC-AGI-1 public evaluation set (ARC Prize model_baseline, pass@2)	1.8% ✎	▃·▇▃▁·█▇▁·▇▃ beats 0/27	very hard · avg 4.1%	▇ 1.0%	sharp · too hard for it · noisy
Icelandic LLM Leaderboard average (Miðeind, Sept 2026 suite)	17.6% ✎	▁▁▃▃▂▄▁▃▄▄▄█ beats 6/38	very hard · avg 14%	▇ 0.9%	typical
SimpleQA Verified	6.0% ✎	▃▁▃▂▃▃▆▆█▃▂▅ beats 0/78	very hard · avg 8.0%	▇ 0.7%	typical
DROP F1, 3-shot (simple-evals) lab	82.2% ✎	▂▂▄▆▆█·▂···▄ beats 9/16	easy · avg 78%	▇ 0.4%	blunt
SimpleQA (simple-evals) lab	7.6% ✎	▄·▁▁▁▁··█▁·▁ beats 0/14	hard · avg 18%	▇ 0.3%	noisy
Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate)	98.0% ✎	▁▁▂▃▄▄█▆▃▂▂▁ beats 122/140	very easy · avg 94%	▇ 0.2%	blunt · maxed out · noisy
ARC-AGI-2 public evaluation set (ARC Prize model_baseline, pass@2)	0.0% ✎	█··▂▅▁▃▁▃·▁▃ beats 0/24	very hard · avg 0.3%	▇ 0.1%	sharp · too hard for it · noisy

$ where to run it

Through an API

api · $0.18/M · $0.1 in / $0.4 out per 1M tokens · 1M context · developers.openai.com ↗ · price sourceUSD per 1M tokens, OpenAI API standard processing, model page 'gpt-4.1-nano' whose default snapshot is gpt-4.1-nano-2025-04-14: input $0.1, output $0.4; cached-input, Batch and other tiers excluded. Currency: USD. OpenAI official model page, observed 2026-09-28.

$ how the score was worked out · OpenAI simple-evals README be… 22%, Epoch AI — internal evaluatio… 21%, Daekeun Kim Korean LLM benchm… 14% and 14 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
OpenAI simple-evals README benchmark results (OpenAI's own models) · lab-reported	7	▇▇▇▇▇▇▇ 22%	MIT
Epoch AI — internal evaluations	4	▇▇▇▇▇▇ 21%	CC BY 4.0
Daekeun Kim Korean LLM benchmark runs	3	▇▇▇▇ 14%	MIT
OpenAI vendor-reported benchmark scores for its own models (openai.com announcement posts and Deployment Safety Hub system cards) · lab-reported	3	▇▇▇ 9%	Factual vendor-reported score (not licensed content)
LMArena leaderboard dataset	2	▇▇ 8%	CC BY 4.0
Aider polyglot benchmark runs	1	▇▇ 6%	Apache-2.0
Arena-Hard-Auto v2.0-Preview leaderboard (repository README)	2	▇▇ 5%	Apache-2.0
Berkeley Function Calling Leaderboard result archive	2	▇ 3%	Apache-2.0
EQ-Bench 3 canonical leaderboard results	1	▇ 3%	MIT
Inception Arabic/Hindi leaderboards (AraGen-v3, HindiGen-v1, Arabic IFEval)	1	▇ 2%	Apache-2.0
Open Persian LLM Leaderboard v2 (Part DP AI and AUT NLP Lab)	1	▇ 2%	Apache-2.0
Galileo Agent Leaderboard (v1 and v2 results)	2	▇ 2%	Apache-2.0
Finance Agent Benchmark (Vals AI, arXiv:2508.00828v1)	1	▇ 1%	CC-BY-4.0
ARC Prize Foundation model_baseline outputs on ARC-AGI-1/2 public evaluation sets	2	▇ 1%	MIT
Miðeind Icelandic LLM Leaderboard	1	▇ 1%	CC BY 4.0
KlinikeBench: Evaluating Language Models Beyond Diagnostic Accuracy (arXiv:2609.38480v2)	1	▇ 0%	CC BY 4.0
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	1	▇ 0%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 52.8