Skip to content
AI Atlas
Family

Llama

All 12 models in the terminal →

first seen 12 Sept 2026

family_01M2A743YF4GR658035W7DCBKH

Models
12
First release
9 Aug 2024
Latest release
17 Jul 2026
Parameters
68M – 8.03B
Artifacts
0
Providers
3

Releases

Family releases by statusActiveDeprecatedOct 24Jan 25Apr 25Jul 25Oct 25Jan 26Apr 26Jul 26fireworks-ai/llama-68m-test — 9 Aug 2024R1 Distill Llama 70B — 23 Jan 2025Salesforce/Llama-xLAM-2-8b-fc-r — 27 Mar 2025Llama Guard 4 12B — 30 Apr 2025nvidia/llama-nemotron-rerank-1b-v2 — 16 Oct 2025perplexity-ai/llama-cpp-linux-binaries — 17 Jul 2026

6 dated releases · dot size = best benchmark rank of the member · 6 members without a sourced release date not shown

Members 12

Family members
ModelParamsContextOpennessLicenceReleasedStatusBest rankCompare
perplexity-ai/llama-cpp-linux-binariesPerplexity AI——Open weights—17 Jul 2026Active—
nvidia/llama-nemotron-rerank-1b-v2NVIDIA1.24B—Open weightsOther16 Oct 2025Active—
Llama Guard 4 12BMeta AI—163.8KOpen weights—30 Apr 2025Active—
Salesforce/Llama-xLAM-2-8b-fc-rSalesforce8.03B—Restricted weightsCC-BY-NC-4.027 Mar 2025Active—
R1 Distill Llama 70BDeepSeek—8.19KOpen weights—23 Jan 2025Deprecated#239 τ²-bench
fireworks-ai/llama-68m-testFireworks AI68M—Open weights—9 Aug 2024Active—
deepseek-r1-distill-llama-8bDeepSeek—128KOpen weights——Deprecated#330 IFBench
Llama Prompt Guard 2 22MMeta AI—————Active—
llama-65bMeta AI—2.05KOpen weights——Deprecated#452 AA Intelligence Index
llama-nemotron-super-49b-v1-5NVIDIA—128KOpen weights——Active#189 GPQA Diamond
SWE-Llama 13BMeta AI—————Active#4 SWE-bench (full test s
SWE-Llama 7BMeta AI—————Active#4 SWE-bench (full test s

Lineage among members 0 relations

Open in Graph →

No lineage relation recorded among these members.

Benchmark progress

Benchmark progress needs at least two dated members with a rank on the same benchmark. Ranks (not scores) are what the family endpoint provides.

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 →