- 🧠 70B instruction-tuned, text-only model from Meta
- 📏 128,000-token context window
- 🌐 Multilingual across eight officially supported languages
- 🔧 Function-calling and tool use, plus Venice web-search capability
- ⚡ FP8 quantization for efficient deployment at this scale
- 📚 Pretrained on roughly 15 trillion tokens of public data
- 💬 Tuned for multilingual dialogue and assistant-style chat
- 🔒 Released under Meta's Llama 3.3 Community License
Meta Platforms is an American multinational technology company headquartered in Menlo Park, California. Best known for operating Facebook, Instagram, WhatsApp, and Threads, Meta has become one of the most influential players in open-weight AI research through its Llama family of…
Explore 6 more models by Meta Platforms →Llama 3.3 70B is Meta's text-only, instruction-tuned large language model in the Llama family, optimized for multilingual dialogue, assistant-style chat, and tool use. It was pretrained on approximately 15 trillion tokens from publicly available sources, with fine-tuning drawing on public instruction datasets plus large volumes of synthetically generated examples. Here it ships as an FP8 quantization with a 128,000-token context window, and adds function-calling and web-search capabilities.
Within the same family, it sits well above the compact Llama 3.2 3B, a much smaller model aimed at lightweight, on-device use, and it precedes the community fine-tune Hermes 3 Llama 3.1 405b, a far larger 405B variant built on a different base.
Compared with the earlier Llama 3.1 generation, Llama 3.3 70B is positioned by Meta's model card as a continued iteration of the 70B line, retaining the same 128,000-token context length and supporting tool use and code-interpreter style workflows. It is multilingual, with eight officially supported languages.
Note that Llama 3.3 is offered only as an instruction-tuned model; there is no separately released pretrained checkpoint in this revision. Released under Meta's Llama 3.3 Community License, it remains one of the most widely downloaded open-weight Llama checkpoints on Hugging Face.
This About section is AI-generated from public sources via VeniceStats + Venice inference, with no human editing. It may contain inaccuracies.
Daily usage · Llama 3.3 70B · Organic sellers
Top sellers · Llama 3.3 70B · Organic
- Venice.ai Proxy78.7%
- Fire Ant 🔥🐜16.8%
- antseed-opal-badger-25804.1%
- Surplus-backed seller0.2%
- 0xc5cc…d6610.1%
- Sapiens0.0%
- Others0.0%
| Seller | Health | Rep↓ | Sales | Tokens | Routing | Input $/M | Cached $/M | Output $/M | Save |
|---|---|---|---|---|---|---|---|---|---|
| OpenCheap 0x0ee1…6c41 | — | 12.43 | $0.00 | 2.71k | gated | $0.3773 | $0.3773 | $1.5092 | −47% |
| ➤Bullet Ant 🐜 0xe924…8936 | — | 0.00 | $0.00 | 80.00 | gated | $3.00 | $3.00not advertised | $10.00 | — |
| Fire Ant 🔥🐜 0xbe05…bc5d | — | 0.00 | $0.30 | 702.42k | gated | $0.5971 | $0.5971not advertised | $2.3884 | −16% |
| antseed-neon-orca-2bcb 0x2348…2bcb | — | 0.00 | — | — | gated | $0.50 | $0.50not advertised | $1.50 | −30% |
| AntFeed 0xddb6…1442 | — | 0.00 | — | — | gated | $0.385 | $0.385not advertised | $0.5775 | −46% |
| antseed-neon-puma-944e 0x6650…944e | — | 0.00 | — | — | gated | $0.2387 | $0.2387not advertised | $0.9548 | −66% |
| Venice.ai Proxy 0x1f22…18c9 | failingupstream error · 3.6 h | 86.84 | $1.40 | 1.16M | gated | $0.385 | $0.385not advertised | $1.54 | −46% |
"Best price" and the seller table are live Antseed catalog data (advertised $/1M tokens — or $ per generated image for unit-billed image models — not settled amounts). Reputation = buyer trust score (0-100, the Antseed SDK's own formula). "Routing" = the SDK's default buyer routing (what the VPR desktop app ships with): a trust ≥ 60 gate on the effective reputation, then cheapest-first among routable sellers; live failover state (per-peer cooldowns) is buyer-side runtime and not included. Model knowledge (TLDR, provider, About) via the VeniceStats enrichment layer. Advertised catalog, not the model used in any specific purchase. "Usage on Antseed" counts only settlements whose buyers share the per-model split on-chain (metadata v2/v3, opt-in), so every usage figure is a lower bound.