DSPy 3.5.0 released
A major release that makes DSPy lighter and faster by replacing heavy dependencies with a minimal LM engine, graduating experimental modules, and adding image support to RLM.
What changed in 3.5.0
- LiteLLM and OpenAI Python library are now optional; DSPy uses a lightweight lm15 engine by default, reducing install size from 206.7 MiB to 30.9 MiB and import time from 1.01s to 0.49s.
- RLM, Flex and GEPA are no longer experimental; ReActV2 is now dspy.ReAct with structured History and typed submit tool.
- XMLAdapter is now the default, changing prompt formatting and output parsing; model_type="auto" is the default for automatic API-family routing.
- RLM now supports image input and can crop images with Pillow; Flex and RLM interpreters run real DSPy with precompiled standard library and cached startup snapshots, reducing startup from 1.59s to 0.45s.
- Disk caches are secure by default with restricted loading; native reasoning is preferred over explicit plans.
Check before you update
- Prompts, defaults and APIs have changed; re-evaluate optimized programs and read the upgrade guide.
- XMLAdapter is now the default Adapter, changing prompts and output contract.
- ReActV2 replaces ReAct; old ReAct name is deprecated until 3.6.
Summarised by AgentGid from Stanford NLP's official release notes on GitHub, as reported there; read the original before relying on it. Price, status and release dates come from our daily data.
DSPy at a glance
| What it is | A framework for programming language models with declarative modules and optimizers instead of hand-written prompts. |
|---|---|
| Latest version | 3.5.0 is the latest stable release we track. |
| Release pace | 1 stable release in the last 30 days; 3.5.0 came 29 days after the previous one |
| Price | Free (OSS). DSPy is open source and free to use; you pay only for the models and infrastructure it runs on. (checked Nov 4, 2026) |
| Downloads, last 7 days | 1.5M up 24.3% |
| Gid Score | 54 #12 in Agent frameworks and SDKs |