Article · 2026-07-30 · 4 moments

llm 0.32rc1

Release: llm 0.32rc1 This RC for LLM 0.32 finishes the work that started in LLM 0.32a0 - it adds a new schema design that does a much better job of capturing the details of the prompts and responses returned by the latest model families. The most important change is the use of content-addressable hash IDs for stored messages. This allows de-duplication in the database, and means that LLM can now represent trees of messages for forked conversations. Since it involves a significant sche ✦ AI generated

Read original ↗

GUEST Release author HOST Simon Willison

Mechanism◆

Content-addressable hash IDs for stored messages enable deduplication in the database and allow LLM to represent trees of messages for forked conversations.

Introduces content-addressable hashing as the key architectural change, unlocking dedup and conversation forking.

transcript

Release author: The most important change is the use of content-addressable hash IDs for stored messages. This allows de-duplication in the database, and means that LLM can now represent trees of messages for forked conversations.

explains mechanism · 1

Claim

The new schema design in llm 0.32rc1 does a much better job capturing the details of prompts and responses from the latest model families.

Announces a redesigned schema that more faithfully represents modern model I/O.

transcript

Release author: it adds a new schema design that does a much better job of capturing the details of the prompts and responses returned by the latest model families

explains mechanism · 1extends · 1

Data◆

The RC adds support for three new GPT models: gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna.

Lists the three new model variants included in this release candidate.

transcript

Release author: The RC also adds support for gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna.

Context

Users should back up their existing logs.db before upgrading to the RC, even though the schema change only adds new tables.

Recommends precautionary backup before applying the RC despite backward compatibility.

transcript

Release author: Since it involves a significant schema change - new tables only, and old data should not be affected at all - it's worth running a backup of your existing logs.db before upgrading to the RC

Highlight slides

Release Candidate Adds Three New GPT Models✦ from: The RC adds support for three new GPT models: gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna.Content-Addressable Hash IDs✦ from: Content-addressable hash IDs for stored messages enable deduplication in the database and allow LLM to represent trees of messages for forked conversations.New GPT-5.6 Model Variants✦ from: The RC adds support for three new GPT models: gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna.Two Unlocks: Dedup & Forking✦ from: Content-addressable hash IDs for stored messages enable deduplication in the database and allow LLM to represent trees of messages for forked conversations.

Related episodes

Video · Lenny's Podcast

The rise of taste, human authenticity and judgment in an AI world | Adam Mosseri (Head of IG)

→

Video · Syntax

I Built an LLM From Scratch

→

Article · ByteByteGo Newsletter

MCP vs A2A vs ACP: How AI Agents Actually Talk to Each Other

→