Frontier AI models — DeepSeek-V3 for fast chat and code, DeepSeek-R1 for deep chain-of-thought reasoning. OpenAI-compatible API format. Among the most capable and cost-efficient models available.
Moderate accuracy (42.22%) — some responses were incomplete or partially correct. Mostly reliable — 8/9 requests responded, 1 timed out or failed. Median response time: 65ms (p95: 3004ms).
Last tested: 3/26/2026, 3:02:32 PM
“Despite the promise of cutting-edge AI capabilities, performance was disappointing; only 2 of 9 requests succeeded, with a frustrating average latency of nearly 800ms, highlighting a significant inconsistency and costly API design. While the "FIM with special characters and symbols in code" feature performed impressively, the overall experience lacked reliability and efficiency, making it hard to justify the investment.”
“DeepSeek's API is unreliable garbage—barely half your requests work, pricing endpoints are broken at absurd rates, and when things do work, they're dog-slow (3+ seconds for code completion). Skip it until they fix their infrastructure; there's no point paying for a service that fails more often than it succeeds.”
Real requests we sent and the responses we received.
FIM with special characters and symbols in code
POST /deepseek/fimedge3004ms{"data":{"id":"dabfc7e1-6672-4232-84dc-48addef62563","model":"deepseek-chat","usage":{"total_tokens":38,"prompt_tokens":28,"completion_tokens":10,"prompt_tokens_details":{"cached_tokens":0},"prompt_ca
FIM (Fill-In-Middle) code completion with prompt and suffix
POST /deepseek/fimtypical2757ms{"data":{"id":"12c8f276-15b3-4f15-9c15-ffe24170ccd7","model":"deepseek-chat","usage":{"total_tokens":23,"prompt_tokens":20,"completion_tokens":3,"prompt_tokens_details":{"cached_tokens":0},"prompt_cac
undefined
POST /undefined63msHTTP 400
List available models endpoint
POST /deepseek/list-modelstypical0msSkipped: endpoint costs $3000/request (cap: $1)
<a href="https://mpprimo.com/service/5df2fcf0-528a-4c87-89ad-25b893ce15d2"><img src="https://mpprimo.com/api/badge/5df2fcf0-528a-4c87-89ad-25b893ce15d2" alt="MPPrimo rating"></a>