Repository navigation
feat(llm): Support GPT-6.1 Sol and Claude Sonnet 5.5 - #1213
Merged
Merged
Conversation
JeanMertz
added this pull request to stack #1215
September 30, 2026 10:57
JP can now query `openai/gpt-6.1-sol` and `anthropic/claude-sonnet-5-5` with their documented capabilities, and `jp init` suggests `claude-sonnet-5-5` as the default Anthropic model. A ChatGPT subscription lists GPT-6.1 Sol alongside the other GPT-6 models. GPT-6.1 Sol accepts `low` through `max` effort but neither `none` nor `minimal`, so it is marked always-on: `reasoning = off` sends the lowest effort instead of a value the API rejects. Its docs limit tool calls to the Responses API, which is the only API JP uses for OpenAI. Sonnet 5.5 rejects `thinking: disabled`. With `reasoning = off`, JP sends `thinking: between_tools`, the lowest setting it accepts, which drops up-front thinking but keeps the short progress updates the model writes between tool calls. Omitting the field instead would run full adaptive thinking. Sonnet 5.5 also rejects a forced `tool_choice` whatever the thinking setting, so a forced tool call becomes `auto` plus an escalating nudge, as on Opus 5.5. `between_tools` passes through the Claude Code (ACP) route as well. The workspace `sonnet`, `claude` and `anthropic` aliases point to Sonnet 5.5, and `sol`, `openai`, `chatgpt` and `gpt` point to GPT-6.1 Sol. Ticket T-0q1k0j4 records how Sonnet 5.5's thinking-block binding changes the options for predicting its 400s. Signed-off-by: Jean Mertz <git@jeanmertz.com>
JeanMertz
force-pushed
the
new-models
branch
from
September 30, 2026 12:26
10c4b35 to
f59b55a
Compare
Signed-off-by: Jean Mertz <git@jeanmertz.com>
Signed-off-by: Jean Mertz <git@jeanmertz.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
JP can now query
openai/gpt-6.1-solandanthropic/claude-sonnet-5-5with their documented capabilities, andjp initsuggestsclaude-sonnet-5-5as the default Anthropic model. A ChatGPT subscription lists GPT-6.1 Sol alongside the other GPT-6 models.GPT-6.1 Sol accepts
lowthroughmaxeffort but neithernonenorminimal, so it is marked always-on:reasoning = offsends the lowest effort instead of a value the API rejects. Its docs limit tool calls to the Responses API, which is the only API JP uses for OpenAI.Sonnet 5.5 rejects
thinking: disabled. Withreasoning = off, JP sendsthinking: between_tools, the lowest setting it accepts, which drops up-front thinking but keeps the short progress updates the model writes between tool calls. Omitting the field instead would run full adaptive thinking. Sonnet 5.5 also rejects a forcedtool_choicewhatever the thinking setting, so a forced tool call becomesautoplus an escalating nudge, as on Opus 5.5.between_toolspasses through the Claude Code (ACP) route as well.The workspace
sonnet,claudeandanthropicaliases point to Sonnet 5.5, andsol,openai,chatgptandgptpoint to GPT-6.1 Sol. Ticket T-0q1k0j4 records how Sonnet 5.5's thinking-block binding changes the options for predicting its 400s.