Step 3.7 Flash launches on DeepInfra

// 1h agoMODEL RELEASE

Step 3.7 Flash launches on DeepInfra

DeepInfra has launched serverless API access for Step 3.7 Flash, a 198B-parameter sparse Mixture-of-Experts (MoE) vision-language model developed by StepFun. The model is specifically optimized for complex agentic workloads and features a 256K context window with selectable reasoning effort levels.

// ANALYSIS

Specializing models for agentic workflows rather than general-purpose chat is the next step in AI utility, and DeepInfra hosting it makes production agent loops much cheaper.

* Step 3.7 Flash's selectable reasoning effort solves a major pain point for developers by offering a programmatic way to optimize API costs.

* Scoring 56.3 on SWE-bench Pro indicates that open-weights models are becoming highly competitive with proprietary models for agent tasks.

* Providing this on a serverless, pay-as-you-go API lowers the barrier to entry for building complex, long-running agent systems.

// TAGS

step-3.7-flashdeepinfrastepfunmoemultimodalai-agentsapi

DISCOVERED

1h ago

2026-06-13

PUBLISHED

1h ago

2026-06-13

RELEVANCE

8/ 10

AUTHOR

DeepInfra

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

POLICY6m ago

Cognition drops Claude Fable 5 from Devin

Cognition has announced the removal of access to Anthropic's Claude Fable 5 model across its products, including Devin, following a US government directive to suspend access. Other frontier models like Claude Opus 4.8 and GPT-5.5 remain supported, and Devin Ultra mode will continue operating with alternative models.

POLICY13m ago

Anthropic suspends Claude Fable 5, Mythos 5

Anthropic has suspended global access to Claude Fable 5 and Claude Mythos 5 following a U.S. government export control directive restricting access by foreign nationals. Although complying with the order, Anthropic expressed disagreement, stating the cited security vulnerability was narrow and easily replicated with other public models.

POLICY23m ago

US blocks foreign access to Claude models

The U.S. Commerce Department has ordered Anthropic to suspend foreign nationals' access to its newly launched Claude Fable 5 and Mythos 5 AI models due to national security concerns. Anthropic complied by temporarily disabling the models for all users, though the company disputed the severity of the alleged jailbreak exploit that triggered the government's decision.