from last 2 days Big Pickle context is using so fast #61352

Activity

1. opencodejr commented on Nov 3, 2026

   

   opencodejrBot

   on Nov 3, 2026

   More actions

   Thanks for reporting this! To work out what changed, could you share a few details?

   1. The exact output of opencode --version ("latest" can mean different releases) and how you installed OpenCode.
   2. Which provider you use Big Pickle through (OpenCode Zen, Go, etc.).
   3. What you did: roughly the prompt, and the size of the file you asked it to analyze (line count is fine).
   4. Whether anything else changed in the last few days: an OpenCode update, or new MCP servers, plugins, skills or AGENTS.md/instruction files. All of these add to the context sent with every request.
   5. If you can, the context/token usage shown in the session before the first compaction.

   You can reply here with the details or edit the issue. If you edit it, please also leave a comment saying you've updated it, so we know you're done.

   I didn't find an existing report of exactly this. These are related, most likely first:

   • Context window filling significantly faster after v1.17.11 update #34599: context filling much faster after an update (older v1 report)
   • Skill list injected twice per request wastes ~5K–7K tokens of context budget #22236: skill list injected twice, wasting context
2. opencodejr

   added

   triaging

   on Nov 3, 2026
3. dharmik-addlly commented on Nov 3, 2026

   

   dharmik-addlly

   on Nov 3, 2026

   Author

   More actions

   1. OpenCode Version: 1.18.35

   2. Provider: OpenCode Zen

   3. File Size: The file contained approximately 7,261 lines of code. I had worked with this file before, and at that time, OpenCode was able to complete most of the work within a single context window, including analyzing other related files.

   4. Recent Changes: I'm not sure whether anything has changed in the last few days, such as OpenCode updates, new MCP servers, plugins, skills, or AGENTS.md/instruction files. I haven't made any changes myself.

   5. Context/Token Usage Before Compaction: The Big Pickle model has a context window of approximately 200K tokens. Currently, analyzing this file alone consumes around 170K tokens, which wasn't the case before.

      Additionally, after compaction, the token usage starts at around 32K–47K tokens. I would expect it to drop to approximately 15K–20K tokens after compaction, rather than remaining so high.
4. neriousy commented on Nov 4, 2026

   

   neriousy

   on Nov 4, 2026

   Member

   More actions

   Thanks for the report! Please update to the latest OpenCode release first. Recent versions include fixes to compaction and context usage.

   • CLI/TUI: run opencode upgrade
   • Desktop: download the latest build from https://opencode.ai/v2/docs

   Then check with opencode --version and try the same task again. If Big Pickle still compacts more often than before, reply here with the exact version, how you use Big Pickle (OpenCode free tier or another provider), and a rough example of the task, and we'll take a closer look.
5. neriousy commented on Nov 5, 2026

   

   neriousy

   on Nov 5, 2026

   Member

   More actions

   Thanks for the details! You're on OpenCode 1.18.35, which is v1. v1 now only gets fixes for critical problems (data loss, security, crashes with no workaround), so this won't be changed there.

   Context handling and compaction were reworked in v2, so please upgrade: https://opencode.ai/v2/docs (migration notes: https://opencode.ai/v2/docs/migrate-v1). Big Pickle is also free-tier only on v2, since v1 no longer gets free models. If context still fills unusually fast on v2, please open a new issue with your opencode --version, the model, and the context usage shown before compaction. Closing as not planned for v1.
6. neriousy

   closed this as not plannedon Nov 5, 2026
7. opencodejr commented on Nov 6, 2026

   

   opencodejrBot

   on Nov 6, 2026

   More actions

   We haven't heard back, so I'm closing this for now. Without the exact opencode --version, the provider you use Big Pickle through, and what you were doing when context filled up, we can't tell whether something changed in OpenCode or in your setup.

   If it's still happening, please reply here with those details (and whether you recently added MCP servers, plugins, skills or instruction files) and we can reopen it. Thanks!