OpenCode Go adds $30 usage credit to its $10 plan

• OpenCode Go now costs $10 per month after a $5 first month and began offering $30 of usage under its first response to the DeepSeek price shock, according to opencode's Phase 1 post.
• DeepSeek's V4 pricing change introduced peak and off-peak billing, with increases ranging from 50% to 1,100% depending on model, token type, and time, opencode's pricing update said.
• OpenCode is sourcing capacity outside China, aggregating provider volume for discounts, and keeping a small amount of DeepSeek traffic as fallback, thdxr's capacity reply said.
• The current Go docs define the allowance as $12 per five hours, $30 per week, and $60 per month, with request counts varying by model, per .
• A quota reset would cost OpenCode an estimated $1 million to $2 million, leaving that option off the table, thdxr's reply said.

OpenCode's updated Go documentation includes a 19-model catalog, model-specific request estimates, and API endpoints alongside the subscription price. DeepSeek's official release note gives the pricing switchover time, while its live pricing table shows the peak and off-peak rates now flowing into the provider economics.

OpenCode announced a $30 usage allowance for Go subscribers paying $10 per month, with the first month priced at $5. The post called that “Phase 1” and said a second phase was starting, while the company was still working through capacity and pricing changes.

The allowance is measured in usage dollars, not a flat request count. OpenCode's current docs list three rolling limits:

• $12 of usage per five-hour window
• $30 of usage per week
• $60 of usage per month

Request volume depends heavily on the selected model. The published monthly estimates range from about 490 Kimi K3 requests and 5,200 DeepSeek V4 Pro requests to 37,800 DeepSeek V4 Flash requests. Go remains an optional provider in OpenCode, connected through /connect, with model IDs formatted as opencode-go/<model>.

DeepSeek launched V4-Pro GA with peak and off-peak API pricing. Its release note says the new rates took effect at 16:00 UTC on August 16, 2026, with off-peak rates set at half the peak rate. Reuters reported that the increases ranged from 50% to 1,100% over the previous prices.

OpenCode confirmed that DeepSeek had raised its prices and that Go limits had been updated in its pricing update.

The current DeepSeek table puts the V4 Pro cache-miss input price at $0.66 per million tokens off-peak and $1.32 at peak. Output costs $1.98 off-peak and $3.96 at peak. V4 Flash is cheaper, but follows the same schedule: $0.22 and $0.44 for cache-miss input, and $0.66 and $1.32 for output.

Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC. That schedule makes time of day part of the cost model for any gateway routing DeepSeek traffic directly.

OpenCode's operator, thdxr, said the team and its partners spent the weekend trying to secure lower-cost capacity, while stressing that OpenCode did not control DeepSeek's original price increase in a follow-up.

The sourcing plan has several parts:

• New capacity is outside China; a small amount of traffic still goes to DeepSeek as a fallback, thdxr said.
• OpenCode pools volume to negotiate provider discounts, then passes those savings through, thdxr explained.
• The team was testing multiple hosting setups because many providers claimed to match DeepSeek's old pricing without actually doing so, thdxr wrote.
• The required hardware layout and physical infrastructure make cheaper inference a capacity problem as much as a pricing problem, thdxr added.

The current Go docs describe the same model more generally: OpenCode says most discounts come from bulk purchasing and reserved GPU capacity, with the target of returning roughly six times the $10 subscription in usage for models where those discounts are available.

OpenCode rejected a blanket quota reset. thdxr estimated that resetting all Go usage would cost between $1 million and $2 million, according to the operator's reply.

The operator also described why the economics are fragile: provider discounts are negotiated on top of sticker prices, and a vendor price change forces OpenCode to rebuild those discounts from scratch thdxr said. Cache reads are a major cost center in that calculation, thdxr added.

A separate OpenCode GitHub issue filed on August 15 reported a user reaching the monthly limit after roughly $32 of visible usage, despite the documentation listing a $60 monthly cap. The issue was automatically closed for missing reproduction details and contains no maintainer explanation for the discrepancy.