EcoHash
An inference service provider offering large models with an OpenAI-compatible API format, supporting streaming completions for open-source models such as llama, with an API base URL of https://api.ecohash.com/v1.
Timeline (last 90 days)
EcoHash's RTX Pro 6000 page explicitly states that the card is a 96GB-per-card Blackwell Server Edition, not an RTX A6000 or RTX 6000 Ada, priced at $1.89 per GPU-hour with per-second billing and a $1 sign-up credit.
The bridge is located in the integrations/retell directory of ecohash-examples, uses the llama-3.1-8b-instruct model by default, and is priced at $0.10 / 1M tokens for both input and output.
The measured performance of llama-3.1-8b-instruct on EcoHash is a TTFT of 205 ms (p95) and about 143 tokens/second for a single stream.
During writing, a local WebSocket test loop was actually run; the bridge completed the config handshake, spoke the first sentence, and streamed back a coherent 18 chunks from llama-3.1-8b-instruct, and ping and interruption handling both worked as documented.
When using llama-3.1-8b-instruct bridged to EcoHash, calculated at about 2000 input tokens and 250 output tokens per minute for a typical customer service call at a unit price of $0.10/1M, the token cost is about $0.0003/minute.
All Facts (5)
EcoHash's RTX Pro 6000 page explicitly states that the card is a 96GB-per-card Blackwell Server Edition, not an RTX A6000 or RTX 6000 Ada, priced at $1.89 per GPU-hour with per-second billing and a $1 sign-up credit.
50%UnverifiedThe bridge is located in the integrations/retell directory of ecohash-examples, uses the llama-3.1-8b-instruct model by default, and is priced at $0.10 / 1M tokens for both input and output.
50%UnverifiedThe measured performance of llama-3.1-8b-instruct on EcoHash is a TTFT of 205 ms (p95) and about 143 tokens/second for a single stream.
50%UnverifiedWhen using llama-3.1-8b-instruct bridged to EcoHash, calculated at about 2000 input tokens and 250 output tokens per minute for a typical customer service call at a unit price of $0.10/1M, the token cost is about $0.0003/minute.
50%UnverifiedDuring writing, a local WebSocket test loop was actually run; the bridge completed the config handshake, spoke the first sentence, and streamed back a coherent 18 chunks from llama-3.1-8b-instruct, and ping and interruption handling both worked as documented.
50%