Opus 4.7 hits regressions, stealth price hike

// 45d agoPRODUCT UPDATE

Opus 4.7 hits regressions, stealth price hike

Anthropic's release of Claude Opus 4.7 sparked developer backlash after replacing the reliable `budget_tokens` parameter with an opaque "Effort Level" system, triggering widespread 400 errors. The update also features a new tokenizer that effectively increases token costs by 35% alongside reports of increased model "laziness" in complex coding tasks.

// ANALYSIS

Anthropic’s 4.7 update is a stability and transparency failure that trades developer trust for increased margins through stealth price hikes.

–Sudden API removals without deprecation cycles are unacceptable for a flagship enterprise-grade model.
–The 35% token cost increase via a new tokenizer is a deceptive way to raise prices without changing the per-million-token rate.
–Hiding reasoning tokens by default obscures the model's logic and makes it harder for developers to monitor or optimize costs.
–The "laziness" reports indicate a regression in reasoning efficiency that now requires manual "high effort" overrides to mitigate.

// TAGS

claude-opus-4-7llmanthropicai-codingreasoningapi

DISCOVERED

45d ago

2026-04-21

PUBLISHED

45d ago

2026-04-21

RELEVANCE

9/ 10

AUTHOR

theo

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

MODEL25m ago

Gemma 4 12B Assistant accelerates decoding

The Gemma 4 12B IT Assistant checkpoint has caused developer confusion, as it is actually a speculative decoding drafter model rather than a standalone chatbot. When configured in tools like vLLM alongside the primary model, it accelerates inference via Multi-Token Prediction without degrading quality.

BENCHMARK25m ago

Google's Gemma 4 12B model faces criticism for slow local performance and poor multimodal generation on Apple Silicon.

A developer tested Google's Gemma 4 12B model on a MacBook Pro M5 Max with 128GB of unified memory, finding its performance disappointing for a model of its size. Achieving only 44 tokens per second, the developer noted that the model's generation quality was subpar—specifically criticizing a produced lava lamp image—and recommended developers stick with Qwen 3.6 27B or 35B instead.

OPEN SOURCE51m ago

A Windows companion suite for the OpenClaw AI agent platform, featuring a system tray app, a WebSocket shared library, and a PowerToys Run/Command Palette extension.

openclaw-windows-node is a Windows companion suite designed for the self-hosted OpenClaw AI agent platform. It enables deeper integration on Windows systems via three main components: a WinUI 3-based system tray application to manage gateway connections and settings, a shared .NET library (OpenClaw.Shared) that communicates with the central OpenClaw gateway using WebSockets, and a PowerToys Command Palette extension that allows users to trigger agent actions directly from their Windows desktop command interface.

Opus 4.7 hits regressions, stealth price hike