Llama 3.3 70B Instruct vs Devstral 2 123B
meta-llama/llama-3.3-70b-instruct · mistralai/devstral-2-123b
As of November 5, 2026, Devstral 2 123B is the cheaper API on UberLLM: $0.0077 per million input tokens and $0.0384 per million output, versus $0.0145 / $0.0465 for Llama 3.3 70B Instruct, about 20% less when output is 3 tokens for every input token. These are the best prices on the market right now. Both work with one OpenAI-compatible key and UberLLM adds no fee. Choose Devstral 2 123B to cut cost; choose Llama 3.3 70B Instruct when its capabilities fit your task better.
Llama 3.3 70B Instruct vs Devstral 2 123B, side by side
| Spec | Llama 3.3 70B Instruct | Devstral 2 123B |
|---|---|---|
| Price · input / 1M | $0.0145 | $0.0077 |
| Price · output / 1M | $0.0465 | $0.0384 |
| Context length | 128,000 | 256,000 |
| Max output tokens | 4,096 | 32,000 |
| Input modalities | text | text |
| Output modalities | text | text |
| Tools / functions | Yes | Yes |
| % off list | 92.7% off | 98.1% off |
Best prices on UberLLM right now. No UberLLM fee on requests; adding credit by card costs 5.5% + $0.50. Updated November 5, 2026. A price more than 70% below list is what buyers were billed, not an error. Why so cheap?
Llama 3.3 70B Instruct vs Devstral 2 123B — FAQ
Is Llama 3.3 70B Instruct or Devstral 2 123B cheaper?
Devstral 2 123B is cheaper on a blended token mix, by about 20%. A request costs what its provider charged, UberLLM adds no fee, and every request is listed with its cost in Activity.
What's the price difference between Llama 3.3 70B Instruct and Devstral 2 123B?
Llama 3.3 70B Instruct: $0.0145 in / $0.0465 out per 1M tokens. Devstral 2 123B: $0.0077 in / $0.0384 out per 1M tokens. That's roughly a 20% blended difference. Prices are live; UberLLM adds no fee.
Can I use both Llama 3.3 70B Instruct and Devstral 2 123B with one API?
Yes. Both are served through a single OpenAI-compatible endpoint — set the base URL to https://api.uberllm.dev/v1 and your UberLLM key, then switch between them by changing the model id. Anthropic SDKs work via https://api.uberllm.dev/v1/messages.