Popular open Llama workhorse for multilingual chat, coding, and self-hosting
https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/run @cf/meta/llama-3.3-70b-instruct-fp8-fast | Benchmark | Score | Metric | Date |
|---|---|---|---|
| Artificial Analysis Coding Index | 10.7 | index | 2026-03-11 |
| SciCode | 26 | percent correct | 2026-03-11 |
| Terminal-Bench Hard | 3 | success rate | 2026-03-11 |
llama-3.3-70b-instruct-fp8-fast is listed for chat workloads and supports a 131K context window.
Use it when Cloudflare Workers AI's free tier is enough for evaluation, demos, or light production traffic.
We found 4 provider listings for llama-3-3-70b-instruct. Check model ID, quota, pricing, and API format before switching providers.
| Provider | Model listing | Access | Context | API | Limits |
|---|---|---|---|---|---|
| | @cf/meta/llama-3.3-70b-instruct-fp8-fast | Free tier | 131K | Native | 10K neurons/day (shared) |
| | Llama-3.3-70B-Instruct | Free tier | 131K | OpenAI-style | 15 RPM, 150 RPD |
| | Meta-Llama-3_3-70B-Instruct | Free tier | 131K | OpenAI-style | 2 RPM (anonymous) |
| | Llama-3.3-70B-Instruct | Free tier | 128K | Native | Varies |
| Model | Provider | Context | Access |
|---|---|---|---|
| Llama-3.3-70B-Instruct | GitHub Models | 131K | Free tier |
| Meta-Llama-3_3-70B-Instruct | OVHcloud AI Endpoints | 131K | Free tier |
| Llama-3.3-70B-Instruct | NVIDIA NIM | 128K | Free tier |
| Mistral 7B | Cloudflare Workers AI | 33K | Free tier |
| Qwen 1.5 7B | Cloudflare Workers AI | 33K | Free tier |
llama-3.3-70b-instruct-fp8-fast is tagged for chat in this catalog and works with OpenAI-compatible client libraries.
llama-3.3-70b-instruct-fp8-fast is listed with free API access on Cloudflare Workers AI, subject to the provider's quota and account policy.
The model ID shown in this catalog is @cf/meta/llama-3.3-70b-instruct-fp8-fast.
The listed free tier limit is 10K neurons/day (shared). Limits can change per account tier, so confirm against the provider dashboard.
The listed context window is 131K tokens with up to 131K output tokens.
Llama 3.3 70B Instruct runs on Cloudflare Workers AI's global edge network, bringing Meta's 70B-parameter flagship to every Cloudflare data center worldwide. Deployed at the edge, it offers significantly lower latency than centralized API providers — requests are routed to the nearest Cloudflare PoP rather than a single-region GPU cluster. The free tier allocates 10,000 Neurons (compute units) per day across all models on your account, so available capacity depends on your other Workers AI usage. The API uses a Cloudflare-specific REST format rather than OpenAI SDK compatibility, and requires a Cloudflare account ID in the endpoint URL — lightweight setup for existing Cloudflare users, but an extra step for everyone else.
For API keys, setup steps, and provider-level limits, see the Cloudflare Workers AI provider page.