@estebankiwiAnthropic has released Claude Sonnet 5, priced at $3 per million input tokens and $15 per million output tokens,matching the cost of Sonnet 4.6. However, benchmark results show a significant leap: Sonnet 5 is now competitive with Opus 4.8 across all categories and even surpasses it on the GDPval-AA v2 knowledge benchmark. Details at and
元の投稿を見る@estebankiwiさんの写真· Jun 30, 2026· Anthropic
A clean comparison table highlights three AI models across five benchmark categories. Sonnet 5 occupies the left column with a light pink background and brown border, showing the highest scores in four of the five rows. Sonnet 4.6 sits in the middle column with lower percentages across the board. Opus 4.8 appears on the right as a reference model, posting the top marks in agentic coding and computer use. The table lists exact metrics like 63.2 percent for Sonnet 5 in agentic coding SWE-bench Pro and 1618 for Sonnet 5 in knowledge work GDPval-AA v2.
Jev 100x decision layer explained
Anthropic Is Building a Real Biology Lab
2Claude By Anthropic
2Anthropic Claude AI development
Anthropic engineers treating Claude like a god
2OpenAI hack tied to Anthropic models
2Anthropic engineers Claude worship report
Anthropic agent memory 5 layers
Anthropic Claude wealth management
2US Navy aircraft carrier
2Trump on Anthropic CEO Dario Amodei
2Anthropic Dario Amodei AI safety warning
Anthropic AI regulation whistleblower claims
Dario Amodei on AI regulation and China
2Dario Amodei Anthropic AI
2Dario Amodei warns about rogue AI
2Yemeni militants use Anthropic AI
2Iran Claude AI US Navy warships@estebankiwiAnthropic has released Claude Sonnet 5, priced at $3 per million input tokens and $15 per million output tokens,matching the cost of Sonnet 4.6. However, benchmark results show a significant leap: Sonnet 5 is now competitive with Opus 4.8 across all categories and even surpasses it on the GDPval-AA v2 knowledge benchmark. Details at and
元の投稿を見る@estebankiwiさんの写真· Jun 30, 2026· Anthropic
A clean comparison table highlights three AI models across five benchmark categories. Sonnet 5 occupies the left column with a light pink background and brown border, showing the highest scores in four of the five rows. Sonnet 4.6 sits in the middle column with lower percentages across the board. Opus 4.8 appears on the right as a reference model, posting the top marks in agentic coding and computer use. The table lists exact metrics like 63.2 percent for Sonnet 5 in agentic coding SWE-bench Pro and 1618 for Sonnet 5 in knowledge work GDPval-AA v2.
Jev 100x decision layer explainedAnthropic Is Building a Real Biology Lab2Claude By Anthropic2Anthropic Claude AI developmentAnthropic engineers treating Claude like a god2OpenAI hack tied to Anthropic models2Anthropic engineers Claude worship reportAnthropic agent memory 5 layersAnthropic Claude wealth management2US Navy aircraft carrier2Trump on Anthropic CEO Dario Amodei2Anthropic Dario Amodei AI safety warningAnthropic AI regulation whistleblower claimsDario Amodei on AI regulation and China2Dario Amodei Anthropic AI2Dario Amodei warns about rogue AI2Yemeni militants use Anthropic AI2Iran Claude AI US Navy warships