Is Llama 3.1 8B deprecated or retired?

Yes. Llama 3.1 8B is Retired per Meta's own deprecation notice, retired on 2026-08-16. ModelPriceWatch last re-verified this status on 2026-10-19.

Is Llama 3.1 8B deprecated?

Meta · lifecycle status verified Oct 19, 2026 against the provider's own page.

Yes — Llama 3.1 8B is retired as of Aug 16, 2026. Meta no longer serves it on the endpoint we price. The prices below are kept for historical reference only.

Last verified pricing

historical reference — not a buy recommendation

Llama 3.1 8B was last priced at $0.050 per 1M input tokens and $0.080 per 1M output tokens, with a 128K-token context window. Its benchmark average was 34 on the blend we track. Full history stays on the Llama 3.1 8B model page.

Same benchmark score or better, for less

computed from live prices + benchmark averages

Cheaper equivalentsscore ≥ 34 on the benchmark blend, ranked by blended cost ((3×input + 1×output) ÷ 4)

Model	Benchmark	Blended $/M	Input $/M	Output $/M
Llama 3.1 8B Retired	34	$0.060	$0.050	$0.080
Nova Micro Amazon cheapest	40	$0.060	$0.035	$0.140	details →
DeepSeek V4 Flash DeepInfra best value	77.5	$0.090	$0.060	$0.180	details →

"Equivalent" here means the model's benchmark average is at or above Llama 3.1 8B's 34. Benchmarks compress real quality differences — validate on your own workload before switching. Model your exact token mix in the migration calculator.

Migrating off Llama 3.1 8B? The migration calculator compares your exact input/output token mix against every current model we track, with benchmark tradeoffs. Lifecycle data on this page comes from Meta's own page and was last re-verified on Oct 19, 2026. See all retired & deprecated models.