DeepSeek eats 76% of OpenCode tokens
Telemetry from terminal coding agent OpenCode shows DeepSeek V4.1 Flash surging from 33% to over 76% of daily token volume after rivals Muse Spark and MiniMax ended their free tiers. With DeepSeek remaining zero-cost on the platform, developer usage rapidly consolidated behind the subsidized open-weights model.
Developer loyalty in agentic coding vanishes the instant free compute disappears. OpenCode's platform data demonstrates that for terminal coding workflows, developers prioritize zero-cost token volume over brand or marginal quality differences.
- –Free tiers dictate developer defaults: When Muse Spark 1.3 Contributor cut free access on October 17 and MiniMax M3.1 Flash went paid on October 23, DeepSeek's share spiked from 33% to 79%.
- –Coding tasks treat models as commodities: Routine edits, terminal loops, and multi-file refactors burn millions of tokens, making a free model the default choice regardless of proprietary alternatives.
- –DeepSeek's aggressive ecosystem subsidy: Keeping V4.1 Flash free within open-source harnesses lets DeepSeek capture the vast majority of developer mindshare while competitors rush to monetize.
- –Open harness portability punishes paid models: Model-agnostic tools like OpenCode allow users to switch providers instantly, eliminating vendor lock-in and exposing models to fierce pricing elasticity.
DISCOVERED
1h ago
2026-11-13
PUBLISHED
1h ago
2026-11-13
RELEVANCE
AUTHOR
AdamHoltererer