Pricing
Silicon Network and inference API prices
The rate of each model on Silicon Network next to the serverless price hosted inference APIs publish for the same open-weight model, per 1M tokens in USD. Every outside figure carries the day it was read on the provider's own page. Prices change: check the source before relying on one.
What is compared
Silicon Network rates come from the published price list, read when this page was last built (November 14, 2026, at most five minutes before you loaded it), and bill the same for every request. Silicon Network serves the quantized weights each model page lists on Apple Silicon Macs; a provider may serve the same model at another precision, and where it names a variant (an FP8 or Turbo build) the row says so. Batch, cached input, priority tiers and dedicated deployments are left out on both sides. A model a provider does not list with a price has no row for that provider. Public cloud list prices are on a page of their own.
Llama 3.1 8B
Llama 3.1 8B, per 1M tokens
| Provider | Offered as | As of | Input | Output |
|---|---|---|---|---|
| Silicon Network | llama-3.1-8b (4bit) | Live price list | Unavailable right now | |
| DeepInfra | Meta-Llama-3.1-8B-Instruct-Turbo (a Turbo build) | $0.02 | $0.04 | |
- DeepInfra, read November 3, 2026 on deepinfra.com: “Meta-Llama-3.1-8B-Instruct-Turbo: input $0.02, output $0.04 per 1M tokens.”
Llama 3.3 70B
Llama 3.3 70B, per 1M tokens
| Provider | Offered as | As of | Input | Output |
|---|---|---|---|---|
| Silicon Network | llama-3.3-70b (4bit) | Live price list | Unavailable right now | |
| DeepInfra | Llama-3.3-70B-Instruct-Turbo (a Turbo build) | $0.10 | $0.32 | |
- DeepInfra, read November 3, 2026 on deepinfra.com: “Llama-3.3-70B-Instruct-Turbo: input $0.10, output $0.32 per 1M tokens.”
DeepSeek V3.1
DeepSeek V3.1, per 1M tokens
| Provider | Offered as | As of | Input | Output |
|---|---|---|---|---|
| Silicon Network | deepseek-v3.1 (4bit) | Live price list | Unavailable right now | |
| DeepInfra | DeepSeek-V3.1 | $0.25 | $0.95 | |
- DeepInfra, read November 3, 2026 on deepinfra.com: “DeepSeek-V3.1: input $0.25 ($0.13 cached), output $0.95 per 1M tokens.”
DeepSeek V3.2
DeepSeek V3.2, per 1M tokens
| Provider | Offered as | As of | Input | Output |
|---|---|---|---|---|
| Silicon Network | deepseek-v3.2 (4bit) | Live price list | Unavailable right now | |
| DeepInfra | DeepSeek-V3.2 | $0.26 | $0.38 | |
- DeepInfra, read November 3, 2026 on deepinfra.com: “DeepSeek-V3.2: input $0.26 ($0.13 cached), output $0.38 per 1M tokens.”
Qwen3 235B
Qwen3 235B, per 1M tokens
| Provider | Offered as | As of | Input | Output |
|---|---|---|---|---|
| Silicon Network | qwen3-235b (4bit) | Live price list | Unavailable right now | |
| Together AI | Qwen3 235B A22B Instruct 2507 FP8 Throughput (FP8 weights) | $0.20 | $0.60 | |
| DeepInfra | Qwen3-235B-A22B-Instruct-2507 | $0.09 | $0.55 | |
- Together AI, read November 3, 2026 on www.together.ai: “Serverless inference: Qwen3 235B A22B Instruct 2507 FP8 Throughput, input $0.20, output $0.60 per 1M tokens.”
- DeepInfra, read November 3, 2026 on deepinfra.com: “Qwen3-235B-A22B-Instruct-2507: input $0.09, output $0.55 per 1M tokens.”
GPT-OSS 20B
GPT-OSS 20B, per 1M tokens
| Provider | Offered as | As of | Input | Output |
|---|---|---|---|---|
| Silicon Network | gpt-oss-20b (MXFP4-Q8) | Live price list | Unavailable right now | |
| Groq | openai/gpt-oss-20b | $0.075 | $0.30 | |
| DeepInfra | openai/gpt-oss-20b | $0.03 | $0.14 | |
- Groq, read November 3, 2026 on console.groq.com: “Production models: openai/gpt-oss-20b, $0.075 input, $0.30 output per 1M tokens.”
- DeepInfra, read November 3, 2026 on deepinfra.com: “gpt-oss-20b, standard tier: $0.03 per 1M input tokens, $0.14 per 1M output tokens.”
GPT-OSS 120B
GPT-OSS 120B, per 1M tokens
| Provider | Offered as | As of | Input | Output |
|---|---|---|---|---|
| Silicon Network | gpt-oss-120b (MXFP4-Q8) | Live price list | Unavailable right now | |
| Together AI | gpt-oss-120B | $0.15 | $0.60 | |
| Fireworks AI | OpenAI GPT OSS 120B | $0.15 | $0.60 | |
| Groq | openai/gpt-oss-120b | $0.15 | $0.60 | |
| DeepInfra | openai/gpt-oss-120b | $0.037 | $0.17 | |
- Together AI, read November 3, 2026 on www.together.ai: “Serverless inference: gpt-oss-120B, input $0.15, output $0.60 per 1M tokens.”
- Fireworks AI, read November 3, 2026 on fireworks.ai: “Serverless: OpenAI GPT OSS 120B, input $0.15, cached input $0.015, output $0.60 per 1M tokens.”
- Groq, read November 3, 2026 on console.groq.com: “Production models: openai/gpt-oss-120b, $0.15 input, $0.60 output per 1M tokens.”
- DeepInfra, read November 3, 2026 on deepinfra.com: “gpt-oss-120b: $0.037 per 1M input tokens, $0.17 per 1M output tokens.”
Not listed
What each provider was checked for and did not list with a price. Models none of them lists have no comparison here; every model and its rate is on the model pages and the pricing page.
- Together AI, checked November 3, 2026: No serverless price for Llama 3.1 8B, Llama 3.1 70B, Llama 3.3 70B, Llama 3.2 3B, DeepSeek V3.1 or V3.2, Kimi K2 Thinking, GLM 4.7 or gpt-oss-20b (some appear for fine-tuning only).
- Fireworks AI, checked November 3, 2026: No serverless price for the Llama models, DeepSeek V3.1 or V3.2, Qwen3 235B, Kimi K2 Thinking, GLM 4.7 or gpt-oss-20b; the page lists newer versions of several families instead.
- Groq, checked November 3, 2026: llama-3.1-8b-instant and llama-3.3-70b-versatile are listed with "Contact sales" instead of a price; no Kimi K2 or Qwen3 235B in production models.
- DeepInfra, checked November 3, 2026: No price for Llama 3.1 70B, Llama 3.2 3B, Kimi K2 Thinking or GLM 4.7.