DeepInfra

Lock-in risk: higher = more lock-in (worse). Overall already inverts this dimension.

Budget open-model workhorse and one of the cheapest large-catalog hosts. Great for evals and high-volume non-critical traffic; dual-run critical paths on Fireworks/Together.

Pricing notes

As of 2026-09-14: Broad OpenAI-compatible catalog with model-specific token rates (e.g. Llama-3.3-70B Turbo $0.10/$0.32 per 1M in/out; DeepSeek-V4-Flash input $0.09, cached input $0.018 and output $0.18 per 1M (deepinfra.com/pricing as of 2026-10-21; not stored as rows); embeddings from $0.005/MTok). Dedicated GPU rentals (deepinfra.com/pricing, re-checked 2026-10-20): A100 80GB $1.49/GPU-hr, H100 $2.90, H200 $3.49, B200 $4.89, B300 $5.89. Standard/Priority(1.5x)/Flex(0.8x) tiers. No long-term contracts.

Sourced pricing

Contract status ambiguous · input_output_split_not_a_single_rate. Single normalized $/1M: null. Machine list: /api/pricing?provider=deepinfra

• Llama-3.3-70B-Instruct-Turbo · usd_per_million_input_tokens · $0.1 · sourced_token_price
• Llama-3.3-70B-Instruct-Turbo · usd_per_million_output_tokens · $0.32 · sourced_token_price
• bge-base-en-v1.5 · usd_per_million_input_tokens · $0.005 · sourced_token_price
• A100 · usd_per_gpu_hour · $1.49 · unit_mismatch · normalized null
• H100 · usd_per_gpu_hour · $2.9 · unit_mismatch · normalized null
• H200 · usd_per_gpu_hour · $3.49 · unit_mismatch · normalized null
• B200 · usd_per_gpu_hour · $4.89 · unit_mismatch · normalized null
• B300 · usd_per_gpu_hour · $5.89 · unit_mismatch · normalized null