Unverified50% confidenceFactExact time
Running an open-source model with 70 billion parameters requires at least 8 H100s in parallel, representing a hardware investment of $240,000–$320,000
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/24/2026
First Seen
Valid until: 10/22/2026
Sources
DeepSeek 500亿估值融资：开源模型如何逼空OpenAI与Anthropic
bilibiliDeeparchWorks7/10/2026
Related Claims
UnverifiedFor a cluster of 10,000 H100s, the GPU hardware cost alone could reach $300 million to $400 million.75% similarUnverifiedOn an NVIDIA H100 GPU, running a model of approximately 200 billion parameters (Opus scale), the cost of generating 1,000 tokens in a single inference is approximately $0.01–$0.03.72% similarUnverifiedRunning inference services on a 70B-parameter open-source model requires at least 2–4 A100/H100-class GPUs, with cloud rental costs of $5–20 per hour.71% similarUnverifiedFine-tuning a 7-billion-parameter model can be done on a consumer-grade GPU, potentially at a cost of only a few hundred dollars70% similarUnverifiedThe price of a single NVIDIA H100 is approximately $30,000 to $40,000, and training a model of GPT-4's scale is estimated to require thousands of H100s running in parallel for several months.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/607058API
curl https://kongchang.com/api/v1/knowledge/claims/607058MCP
get_claim(id=607058)