whatmodel.sh / GPT-3 text-curie-001

GPT-3 text-curie-001

+ contribute

GPT-3 text-curie-001 is a closed-weights language model from OpenAI, released January 2022. Licence: proprietary.

Catalogue identity notes

OpenAI API InstructGPT-series completion engine text-curie-001. Identity from Wei et al. (arXiv:2201.11903v6): 'for which we use text-ada-001, text-babbage-001, text-curie-001, and text-davinci-002, which presumably correspond to InstructGPT models of 350M, 1.3B, 6.7B, and 175B parameters' (the 6.7B size is the authors' presumption, so params_b is omitted). OpenAI made InstructGPT models the API default on 2022-01-27 ('Aligning Language Models to Follow Instructions', archived). Retired with the other InstructGPT engines on 2024-01-04 (OpenAI deprecations page). Comparison-only.

Model information sources

Release date ↗

12.2±1.4comparison only: no current run option verifiedfrom 18 benchmark results across 9 sources

Score within the LMArena (text) comparison set only.

$ how it compares

The LLMs ranked either side of GPT-3 text-curie-001, out of 877 with a score, then more from OpenAI.

#	model	score	where	RAM	released
852	LaMDA 8B (pre-trained) Google	12±2	never released		2022-01
853	Pythia 1B EleutherAI	12±1	local	~2.2GB	2023-02
854	GPT-3 text-curie-001 OpenAI	12±1	comparison only		2022-01
855	GPT-Neo 1.3B EleutherAI	12±2	local	~7.1GB	2021-03
856	T5 Small Google	12±3	local	~1.8GB	2019-10
	more from OpenAI
1	GPT-6 Astra OpenAI	86±1	api $20.00/M		2026-09
3	GPT-6.1 Sol OpenAI	84±1	api $4.00/M		2026-09
8	GPT-6 Sol OpenAI	81±1	api $4.00/M		2026-09

see all LLMs · find the ones that fit your machine

$ what it's good at

math▁▃▃▆█▇▅▇▃▁▁▁ #506 of 523 * · 5.4% of its score · 1 test

Beats Flan-T5 Large, OPT-66B, Flan-T5 Base here, though they rank higher overall. Trails GPT-Neo 125M, OPT-125M, GPT-2 XL (1.5B) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
GSM8K +1	2.4% ✎	█·▅▃▃▃·▃·▃▃▅ beats 5/12	easy · avg 67%	▇▇ 5.4%	sharp · too hard for it · several runs

other tests16 not in a task · 94% of its score

test	result	vs the others in its run	difficulty	counts	why
MAWPS (CoT-paper union)	3.5% ✎	▅▃█···▅·▃▃▃▅ beats 4/12	very easy · avg 94%	▇▇▇▇▇▇▇▇▇▇▇▇▇ 32%	sharp · too hard for it
ASDiv	3.6% ✎	▅▃▃▅··▃▃·▃▃█ beats 3/12	easy · avg 83%	▇▇▇▇▇▇▇ 17%	sharp · too hard for it
SVAMP	3.1% ✎	▃▅█···▅·▃▃▃▅ beats 4/12	easy · avg 82%	▇▇▇▇▇ 14%	sharp · too hard for it
BIG-bench Date Understanding	8.9% ✎	▃·▅·▅▃▅·▃▃·█ beats 4/12	medium · avg 52%	▇▇▇ 6.3%	too hard for it
CSK-PN boolean commonsense QA (10-shot)	67.9% ✎	▃·▃·····▃▃▃█ beats 0/7	easy · avg 87%	▇▇ 3.9%	blunt · at its level
New Yorker Caption Contest matching, from description, 5-shot (accuracy)	20.4% ✎	█·····▃··▅·▃ beats 0/6	medium · avg 71%	▇▇ 3.8%	sharp · too hard for it
CommonsenseQA	19.0% ✎	█······▁·▁▁▃ beats 3/12	easy · avg 87%	▇ 3.5%	sharp · too hard for it
OpenBookQA	30.6% ✎	▅▁▁▁▃▂▃█▃▁▃▁ beats 30/107	medium · avg 70%	▇ 3.0%	too hard for it · small run

8 more tests

test	result	vs the others in its run	difficulty	counts	why
BIG-bench Sports Understanding	0.0% ✎	█·····▃·▁·▄▁ beats 3/12	medium · avg 81%	▇ 2.5%	sharp · too hard for it
StrategyQA	53.6% ✎	█·····▅▃▃▅▃█ beats 4/12	medium · avg 70%	▇ 2.3%	too hard for it
BoolQ	66.2% ✎	▁····▁▂█▆▅▆▄ beats 42/103	easy · avg 88%	▇ 2.2%	noisy · small run
LMentry accuracy (zero-shot)	41.7% ✎	▂▆·▄▄█▂·▂··▂ beats 7/14	medium · avg 65%	▇ 1.2%	blunt · at its level
ConvRe Re2Text, normal relation, zero-shot (prompt 1#)	53.5% ✎	▂▂▄··▂▂█▄··▂ beats 3/12	easy · avg 90%	▇ 1.1%	noisy
CREPE event likelihood change (3-shot, macro F1)	35.6% ✎	██····█····█ beats 1/3	medium · avg 52%	▇ 1.0%	blunt
AQuA-RAT	13.4% ✎	█▁·▁▁·····▂▁ beats 4/12	medium · avg 48%	▇ 0.5%	sharp · too hard for it
FANToM BeliefQ [Choice], short context (zero-shot)	21.0% ✎	█·····▂····▁ beats 2/11	very hard · avg 55%	▇ 0.0%	sharp · too hard for it · noisy

$ where to run it

No verified API or download recorded yet · add one

$ how the score was worked out · Chain-of-Thought Prompting El… 82%, In-Context Demonstration Sele… 5%, Say What You Mean! Large Lang… 4% and 6 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
Chain-of-Thought Prompting Elicits Reasoning in Large Language Models — authors' own runs	9	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 82%	CC BY 4.0
In-Context Demonstration Selection with Cross Entropy Difference — authors' own runs	2	▇▇ 5%	CC BY 4.0
Say What You Mean! Large Language Models Speak Too Positively about Negative Commonsense Knowledge — authors' own runs	1	▇ 4%	CC BY 4.0
Do Androids Laugh at Electric Sheep? Humor ``Understanding'' Benchmarks from The New Yorker Caption Contest — authors' own runs	1	▇ 4%	CC BY 4.0
LMentry: A Language Model Benchmark of Elementary Language Tasks — authors' own runs	1	▇ 1%	CC BY 4.0
An Investigation of LLMs' Inefficacy in Understanding Converse Relations — authors' own runs	1	▇ 1%	CC BY 4.0
Causal Reasoning of Entities and Events in Procedural Texts — authors' own runs	1	▇ 1%	CC BY 4.0
Bridging the Resource Gap: Exploring the Efficacy of English and Multilingual LLMs for Swedish — authors' own runs	1	▇ 1%	CC BY 4.0
FANToM: A Benchmark for Stress-testing Machine Theory of Mind in Interactions — authors' own runs	1	▇ 0%	CC BY 4.0
prior · pull toward the middle		▇ 1%
weighted average		= 12.2

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated 2026-10-14 08:02 UTC · sources and licences · json