Skip to content
AI Atlas
Familyroot · Llama

Llama 3.1

All 18 models in the terminal →

first seen 11 Sept 2026

family_01M2A55V3VWQ8G1VHPF235PAVT

Models
18
First release
14 Jul 2024
Latest release
28 Aug 2024
Parameters
8.03B – 405.9B
Artifacts
3
Providers
1

Releases

Family releases by statusActiveDeprecatedAug 24Llama 3.1 8B — 14 Jul 2024Llama-3.1-405B — 16 Jul 2024Llama 3.1 70B Instruct — 23 Jul 2024Llama 3.1 8B Instruct — 23 Jul 2024NousResearch/Meta-Llama-3.1-70B-Instruct — 24 Jul 2024NousResearch/Meta-Llama-3.1-8B — 24 Jul 2024NousResearch/Meta-Llama-3.1-8B-Instruct — 24 Jul 2024Hermes-3-Llama-3.1-8B — 28 Jul 2024Llama 3.1 Euryale 70B v2.2 — 28 Aug 2024

9 dated releases · dot size = best benchmark rank of the member · 9 members without a sourced release date not shown

Members 18

Family members
ModelParamsContextOpennessLicenceReleasedStatusBest rankCompare
Llama 3.1 Euryale 70B v2.2Sao10K—131.1KOpen weights—28 Aug 2024Active—
Hermes-3-Llama-3.1-8BNous Research8.03B—Restricted weightsLlama-3-Community28 Jul 2024Active—
NousResearch/Meta-Llama-3.1-70B-InstructNous Research70.5B—Restricted weightsLlama-3.1-Community24 Jul 2024Active—
NousResearch/Meta-Llama-3.1-8BNous Research8.03B—Restricted weightsLlama-3.1-Community24 Jul 2024Active—
NousResearch/Meta-Llama-3.1-8B-InstructNous Research8.03B—Restricted weightsLlama-3.1-Community24 Jul 2024Active—
Llama 3.1 70B InstructMeta AI70.5B131.1KRestricted weightsLlama-3.1-Community23 Jul 2024Active—
Llama 3.1 8B InstructMeta AI8.03B131.1KRestricted weightsLlama-3.1-Community23 Jul 2024Active—
Llama-3.1-405BMeta AI405.9B128KRestricted weightsLlama-3.1-Community16 Jul 2024Active#184 Terminal-Bench
Llama 3.1 8BMeta AI8.03B128KRestricted weightsLlama-3.1-Community14 Jul 2024Deprecated#262 Terminal-Bench
deephermes-3-llama-3-1-8b-previewNous Research—128KOpen weights——Active#360 Humanity's Last Exam
hermes-4-llama-3-1-405bNous Research—128KOpen weights——Active#167 Terminal-Bench
hermes-4-llama-3-1-70bNous Research—128KOpen weights——Active#240 τ²-bench
llama-3-1-nemotron-instruct-70bNVIDIA—128KOpen weights——Active#208 Terminal-Bench
llama-3-1-nemotron-nano-4b-reasoningNVIDIA—128KOpen weights——Deprecated#286 τ²-bench
llama-3-1-nemotron-ultra-253b-v1-reasoningNVIDIA—128KOpen weights——Active#205 GPQA Diamond
Llama-3.1-70BMeta AI—128KOpen weights——Deprecated#231 Terminal-Bench
Meta-Llama-3.1-70BMeta AI——Open weights——Active—
Meta-Llama-3.1-8BMeta AI——Open weights——Active—

Lineage among members 6 relations

Open in Graph →
Lineage of the Llama 3.1 familyLlama 3.1 8B Instruct fine-tuned from Llama 3.1 8BHermes-3-Llama-3.1-8B fine-tuned from Llama 3.1 8BLlama 3.1 8B Instruct derived from Meta-Llama-3.1-8BHermes-3-Llama-3.1-8B derived from Meta-Llama-3.1-8BLlama 3.1 70B Instruct fine-tuned from Llama-3.1-70BLlama 3.1 70B Instruct derived from Meta-Llama-3.1-70BLlama 3.1 8B8.03BLlama 3.1 8BLlama 3.1 70B Instruct70.5BLlama 3.1 70B InstructLlama 3.1 8B Instruct8.03BLlama 3.1 8B InstructHermes-3-Llama-3.1-8B8.03BHermes-3-Llama-3.1-8BLlama-3.1-70BLlama-3.1-70BMeta-Llama-3.1-70BMeta-Llama-3.1-70BMeta-Llama-3.1-8BMeta-Llama-3.1-8B

Benchmark progress

Terminal-Benchbest #184

Best rank of the family on Terminal-Bench#400#300#200#100#0Jul 24Jul 24Jul 24Jul 24

IFBenchbest #208

Best rank of the family on IFBench#400#300#200#100#0Jul 24Jul 24Jul 24Jul 24

τ²-benchbest #257

Best rank of the family on τ²-bench#400#300#200#100#0Jul 24Jul 24Jul 24Jul 24

Humanity's Last Exambest #294

Best rank of the family on Humanity's Last Exam#400#300#200#100#0Jul 24Jul 24Jul 24Jul 24

GPQA Diamondbest #320

Best rank of the family on GPQA Diamond#600#400#200#0Jul 24Jul 24Jul 24Jul 24

Artificial Analysis Intelligence Indexbest #324

Best rank of the family on Artificial Analysis Intelligence Index#400#300#200#100#0Jul 24Jul 24Jul 24Jul 24

Best rank (primary comparability group, current leaderboard) reached by any member released up to each date — lower is better, axis inverted. Ranks are today's ranks, not the ranks at release time.

Aggregates are computed from the members' sourced claims; members without a release date or parameter count are left out of the range and timeline rather than guessed. Methodology →