OpenCode cuts DeepSeek, GLM prices 60%
OpenCode slashed its hosted inference pricing for DeepSeek V4.1 Flash and GLM 5.3 by 60%, undercutting providers like Fireworks and Baseten. The developer platform serves the open-weight models from its own GPU fleet to lower token costs for high-volume agentic coding workflows.
Developer tools are aggressively collapsing inference margins: by hosting open-weight models on in-house GPUs, OpenCode is bypassing API middlemen and daring standalone inference clouds to match its unit economics.
- –DeepSeek V4.1 Flash drops to $0.12 per million input tokens ($0.48 output) and GLM 5.3 to $0.56 input ($1.76 output), drastically reducing iteration overhead for terminal agents.
- –OpenCode explicitly targeted Fireworks AI and Baseten, betting that routing massive developer traffic across proprietary GPU clusters yields superior margins over general-purpose model hosts.
- –Agentic coding loops consume heavy context through iterative tool calls and test executions, making token pricing the defining battleground for developer platform adoption.
- –Vertical integration allows coding assistants to treat model inference as a low-margin utility, placing severe pressure on pure-play inference providers relying on token markups.
DISCOVERED
2h ago
2026-11-21
PUBLISHED
2h ago
2026-11-21
RELEVANCE
AUTHOR
thdxr