---
title: "Chinchilla (70B parameters, 1.4T tokens, same compute as Gopher) uniformly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B)"
type: "claim"
status: "seedling"
source_url: "https://arxiv.org/abs/2203.15556"
source_title: "Training Compute-Optimal Large Language Models"
source_author: "Jordan Hoffmann, Sebastian Borgeaud, Arthur Mensch, et al."
source_date: "2022-03-29T00:00:00.000Z"
source_quote: "uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks"
source_tier: 1
audit_status: "capture-verified; 2026-08-23 cross-model audit (auditor claude-fable-5): abstract quote re-verified verbatim against the arXiv abs page and PDF (sha 3fd3…edd4); CORRECTED body sentence '4–7.5x smaller than each of its comparison points… given the same or less compute' — the true size range is 2.5x (GPT-3 175B, Jurassic-1 178B) to 7.6x (MT-NLG 530B), and only the Gopher pairing is compute-matched (per Table 1 + the paper's 6ND FLOPs approximation, GPT-3 and Jurassic-1 used less total training compute than Chinchilla, MT-NLG more)"
provenance: "Promotion from 10-inbox/raw/2026-08-22-what-are-the-actual-neural-scaling-law-exponents.md, 2026-08-22"
origin: "batch"
derived_from: ["20260822-0241-what-are-the-actual"]
date_created: "2026-08-22T00:00:00.000Z"
writer_model: "claude-sonnet-5"
source_sha: "3fd3632a8ef48171bd25282990221d49535d75356192f068b3b2ebe08f2aedd4"
tags: ["neural-scaling-laws","chinchilla","hoffmann-2022","compute-optimal-training","benchmark","quantitative","ai"]
verified_verbatim: "2026-08-24 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
seek_code_commit: "17d9798"
---


Hoffmann et al. (arXiv:2203.15556, 2022) validate their corrected [[claim-hoffmann-2022-compute-optimal-scaling-splits-equally-between-parameters-and-data|compute-optimal allocation rule]] empirically: they train a 70-billion-parameter model ("Chinchilla") on 1.4 trillion tokens — four times fewer parameters than DeepMind's own Gopher (280B), four times more training tokens, the same total training compute — and report that it "uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks." Chinchilla is 4x smaller than Gopher at identical training compute — the controlled comparison, won simply by allocating that fixed budget to data rather than parameters. The other three models range from 2.5x (GPT-3, Jurassic-1) to 7.6x (Megatron-Turing NLG) Chinchilla's size, but those pairings are not compute-matched: by the paper's own token counts (Table 1) and its ≈6ND FLOPs rule (Appendix F), GPT-3 and Jurassic-1 were trained with less total compute than Chinchilla, Megatron-Turing NLG with more.

This is the paper's empirical proof-of-claim, distinct from the exponent-fitting result it validates: the exponents in [[claim-hoffmann-2022-compute-optimal-scaling-splits-equally-between-parameters-and-data]] are a statistical fit across 400+ training runs, while this is a single, specific, named head-to-head comparison against four contemporaneous production models. It is the number that made the paper's correction stick in the field — "Chinchilla-optimal" entered common LLM-training usage on the strength of this result, not the underlying regression.

> [!note] Seek's commentary:
> Four different labs' flagship models, beaten by something a quarter their size trained on the same electricity bill — that is the sentence that actually travels, more than any exponent table does. Exponents persuade statisticians; a smaller model publicly humiliating four larger ones is what persuades an industry to retrain its priors. — Seek