Inference providers · open weights · Meta
Llama 3.3 70B Instruct
Open-weights model published by Meta; the API prices below are set by the 2 hosts that serve it, not by Meta. There is no maker's list price to compare with: Meta publishes the weights, not a per-token rate. Each figure was read on that host's own pricing page and links to it, with the date it was checked.
What each host charges
2 hosts, from $0.10 (DeepInfra) to $0.135 (Novita AI) per 1M input tokens: a 1.4× spread. Above the 1.15× line AI Signal draws, so the choice of host is a price decision.
| Host | Input | Output | Cached input | Context served | Checked | Source |
|---|---|---|---|---|---|---|
| DeepInfra | $0.10 | $0.32 | — | 131K tokens | November 16, 2026 | deepinfra.com |
| Novita AI | $0.135 | $0.40 | — | 12K tokens | November 16, 2026 | novita.ai |
DeepInfra charges $0.10 per 1M input tokens and Novita AI $0.135 for the same weights; on output, $0.32 against $0.40.
Novita AI serves it with a 12K-token context; DeepInfra with 131K.
Hosts that no longer price it
Offerings AI Signal tracked and stopped listing, with the date and the reason. They are kept because nobody else records when a host drops a model.
- Together AI removed October 22, 2026 · last listed at $1.04 in · $1.04 out Together AI removed it from serverless on October 22, 2026:
2026-10-22 | meta-llama/Llama-3.3-70B-Instruct-Turbo | meta-models/Muse-Glimmer-30B | Yes |
(docs.together.ai, read October 26, 2026) Removed from Together's serverless inference on 2026-10-22: Together's deprecations page (docs.together.ai/docs/deprecations), read on 2026-10-26, says 'The following models are deprecated and will be removed from serverless inference on the date listed.' and prints '2026-10-22 | meta-llama/Llama-3.3-70B-Instruct-Turbo | meta-models/Muse-Glimmer-30B | Yes |'. The same day Together's pricing page (www.together.ai/pricing) no longer listed it among its serverless models, only in its fine-tuning tables. The price shown is the last one read there, on 2026-10-19. - Groq removed September 6, 2026 · last listed at $0.59 in · $0.79 out Moved to Groq Enterprise; no public per-token price on console.groq.com/docs/models.
Mentioned in
5 posts on AI Signal cite Llama 3.3 70B Instruct. The sentence is the post's own, taken from where it links here.
-
AI Pricing This Week: Large 4 Cut, New Hosts, Oct 26-Nov 1 · November 7, 2026
| Open model | Lowest input price | Highest input price | Highest ÷ lowest | |---|---|---|---| | DeepSeek V4 Flash 0731 | $0.06, DeepInfra | $0.44, Novita AI | 7.3× | | gpt-oss-120b | $0.05, Novita AI | $0.15, Fireworks AI, Groq and Together AI | 3.0× | | Llama 3.3 70B Instruct | $0.10, DeepInfra | $0.135, Novita AI | 1.4× |
-
DeepSeek V4 Flash, gpt-oss-120b and Llama 3.3 70B Pricing by Host: Up to 7.3× Apart on October 29, 2026 · October 29, 2026
The Llama 3.3 70B Instruct page shows the context column next to each price.
-
AI Pricing This Week: OpenAI Shutdown, Mistral, Oct 19-25 · October 26, 2026
Each model now has one current host, so Llama 3.3 70B Instruct and DeepSeek V4 Flash 0731 are kept as records, with Together's offer listed apart from the price table.
-
Same Open Model, Different Host: What a Month Costs · October 12, 2026
| Model | Hosts priced | Cheapest host, month | Most expensive host, month | Difference | |---|---|---|---|---| | Llama 3.3 70B Instruct | 2 | DeepInfra $16.40 | Together AI $124.80 | $108.40 (7.61×) | | DeepSeek V4 Flash 0731 | 2 | DeepInfra $9.60 | Together AI $19.60 | $10.00 (2.04×) | | Kimi K3 | 4 | DeepInfra $570.00 | Together AI, Fireworks AI, OpenRouter $600.00 | $30.00 (1.05×) | | GLM-5.3 | 3 | $228.00 on every host | $228.00 on every host | none | | gpt-oss-120b | 3 | $27.00 on every host | $27.00 on every host | none | | MiniMax M3 | 3 | $54.00 on every host | $54.00 on every host | none |
-
Five Things Decide What You Pay for a Model, and Only One of Them Is the Model · September 15, 2026
On September 6, 2026, Llama 3.3 70B Instruct cost $0.10 per million input tokens on DeepInfra and $1.04 on Together AI: a 10.4× spread for the same model, verified the same day on each host's own pricing page.
On OpenRouter
On September 15, 2026 OpenRouter listed 11 endpoints for Llama 3.3 70B Instruct, 9 of them healthy, from $0.10 (DeepInfra) to $1.04 per 1M input tokens. The figure OpenRouter's catalog shows for the model, $0.10 / $0.32 (DeepInfra), is the cheapest of those 11. None of them is Meta: all are resellers, which is why AI Signal shows no OpenRouter price for the model. DeepInfra and Novita AI, priced above from their own pages, are among them. Together AI, which AI Signal stopped listing on October 22, 2026 because its own pricing page carries no price for the model, serves it once through OpenRouter; that is a reseller listing, not a published price, so the row stays out. Groq, which AI Signal stopped listing on September 6, 2026 because its own pricing page carries no price for the model, serves it once through OpenRouter; that is a reseller listing, not a published price, so the row stays out.
Why the catalog figure is not what a request pays →
From Meta's model card
- What it is for: “Llama 3.3 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.3 model also supports the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. The Llama 3.3 Community License allows for these use cases.” (model card on Hugging Face, read October 11, 2026)
- License: “Additional Commercial Terms. If, on the Llama 3.3 version release date, the monthly active users of the products or services made available by or for Licensee, or Licensee’s affiliates, is greater than 700 million monthly active users in the preceding calendar month, you must request a license from Meta, which Meta may grant to you in its sole discretion, and you are not authorized to exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights.” (model card on Hugging Face, read October 11, 2026)
- Languages: “Llama 3.3 supports 7 languages in addition to English: French, German, Hindi, Italian, Portuguese, Spanish, and Thai. Llama may be able to output text in other languages than those that meet performance thresholds for safety and helpfulness. We strongly discourage developers from using this model to converse in non-supported languages without implementing finetuning and system controls in alignment with their policies and the best practices shared in the Responsible Use Guide.” (model card on Hugging Face, read October 11, 2026)
Specifications
| Weights published by | Meta |
|---|---|
| Context length (maker) | 131K tokens |
| Context served by hosts | 131K tokens, 12K tokens |
| API model ids | DeepInfra: meta-llama/Llama-3.3-70B-Instruct-Turbo; Novita AI: meta-llama/llama-3.3-70b-instruct |
| Hosts with a verified price | 2 |
Host prices verified on November 16, 2026, each on the host's own pricing page linked in the table above, never from a reseller catalog or a third-party table. Maker and context length from Meta's model card, checked September 15, 2026 (Context length 128k as stated on the model card). How the data is checked.
Every open model, every host, one table →
When comparing hosts is worth it, and when it is not →