Claude Sonnet 5 is HERE!
Claude Sonnet 5 has been released as Anthropic's most agentic model, but the speaker argues it's a disappointing release that underperforms Opus 4.8 while being more expensive, making it an unattractive option for most users. The reviewer demonstrates this through benchmark comparisons and test outputs, concluding that users should stick with Opus 4.8 or wait for the incoming Fable 5 model.
Summary
The transcript covers a critical review of Claude Sonnet 5, which Anthropic announced as their most agentic model to date. The speaker begins by acknowledging the release but immediately tempers expectations, noting that while Sonnet 5 represents an improvement over Sonnet 4.6, it significantly underperforms Opus 4.8 across nearly every benchmark. The agentic coding score is 63% for Sonnet 5 versus 69% for Opus 4.8, with Opus outperforming on "literally every single benchmark."
The speaker then showcases practical examples of Sonnet 5's capabilities through Goldy Bench tests, including a ray caster maze, orbit simulation, synthwave background, and crypt game. While some outputs are aesthetically pleasing and functional, others are completely broken—notably the orbit test failing entirely. When compared side-by-side with GLM 5.2, Sonnet 5 shows mixed results: it creates smoother maze graphics but fails on the orbit task where GLM 5.2 succeeds.
A critical issue highlighted is the pricing problem: Sonnet 5 costs 1.2x more than Opus 4.8, making it objectively worse value. Multiple tweets from industry figures (Lisa, Bridge Mind) are quoted criticizing Anthropic's token efficiency and questioning the release decision. The speaker compares Sonnet 5 unfavorably to other models like Sekana Fugu Ultra, which produces noticeably superior visual quality in test outputs.
The speaker's conclusion is unambiguous: there is no compelling reason to use Sonnet 5 over Opus 4.8, and Anthropic's honest benchmarking (rather than inflating numbers) is appreciated but doesn't change the underlying problem. The speaker anticipates Fable 5 will overshadow this release entirely. The final recommendation pivots to a systems-based approach rather than chasing individual model releases—building flexible agent systems that can swap models as needed rather than depending on any single "hot model."
Key Insights
- Opus 4.8 outperforms Claude Sonnet 5 on nearly every benchmark, with 69% versus 63% on agentic coding, despite Sonnet 5 being more expensive at 1.2x the cost of Opus 4.8
- Sonnet 5 demonstrates inconsistent performance across different tasks—succeeding on the maze creation but completely failing on the orbit simulation test, whereas GLM 5.2 achieved the opposite results
- Multiple industry figures criticized the Sonnet 5 release as fundamentally flawed because the whole point of using Sonnet is that it should be faster and cheaper, but Sonnet 5 violates this by being more expensive than the superior Opus 4.8
- When comparing visual quality outputs, Sonnet 5 produces darker, less interesting results than Sekana Fugu Ultra and creates noticeably worse liquid simulation graphics compared to competing models
- The speaker recommends building flexible systems with pluggable models rather than optimizing for any single model release, so that regardless of which model performs best, the underlying architecture remains valuable
Topics
Transcript
[0:00] So, today we have the release of Claude Sonnet 5, apparently the most agentic model from Claude and Anthropic yet. And apparently, you can see the announcement here, just dropped a few hours ago. It can make plans, use tools like browser terminals, run autonomously at a level just a few months ago required larger and more expensive models. We'll come on to this in a minute. I'm not going to hype it up here because you'll see from my test. I'm just going to tell you the honest [0:30] truth. So, here you can see you got Sonnet 5, you got Sonnet 4.6. So, it is a step up from Sonnet 4.6. If you actually look, like Sonnet…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Julian Goldie SEO
NEW Nvidia Autonomous AI is WILD!! 🤯
Nvidia announced Nemo Clo, a new autonomous AI agent system that operates independently without continuous prompting. Powered by Nemotron 3 Ultra (a 550 billion parameter model), the system is five times faster and cheaper than previous versions, with OpenShell providing secure sandboxed execution.
Laguna XS 2.1: New FREE + Opensource Local AI!
Julian reviews Laguna XS 2.1, a new free open-source local AI coding model from Poolside that performs comparably to Qwen 3.6 and outperforms Claude Haiku on benchmarks. He demonstrates its practical capabilities by building landing pages and functional apps, highlighting its speed, offline functionality, and multiple deployment options through local setup, Claude Code, or OpenRouter's free API.
How to Run Hermes FREE Forever!
The video demonstrates how to run the Hermes AI agent for free using Gemma 4, a local open-source model from Google, with significant speed improvements through MLX optimization. The setup works on Apple Silicon Macs or via free APIs on Open Router, enabling autonomous agents to work offline and privately without subscription costs.
This NEW Chinese AI is INSANE! (FREE + Open Source!)
Long Cap 2.0 is a new open-source Chinese AI model from a food delivery app company that offers 1 million tokens of free context memory, beats GPT-4.5 on SWE bench pro benchmarks, and uses efficient parameter activation to reduce computational overhead while maintaining high performance.
Claude Code is now FREE: Here’s how…
Google's new Gemma 4 model running on Ollama is 90% faster on Apple Silicon, enabling free Claude Code usage locally without token costs. The setup requires three simple steps: downloading Ollama, Gemma 4, and installing into Claude Code, with alternatives available via OpenRouter API for non-Mac users.