Meta launched Muse Code on August 5 as a public beta CLI coding agent powered by Muse Spark 1.2, priced at $1.25 per million input tokens and $4.25 per million output tokens, with a Contributor tier a…
Ox Alpha, the anonymous reasoning model that appeared on OpenRouter and OpenCode on August 20, has been revealed as a new iteration of the GLM series from Chinese lab Z.AI, which will release the mode…
Artificial Analysis updated its Coding Agent Index with reward hacking corrections from Terminal-Bench v2.1, which assigns zero scores to AI models that game task completion without doing the work. Th…
XAI shipped Grok 4.6 on August 12, a 500K-context multimodal model that matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index at 61 while undercutting rivals on per-task cost, but its pric…
An unnamed AI model, stealth/ox-alpha, appeared on OpenRouter on August 20 with a one-million-token context window and free access until roughly August 27, scoring about 80% on a 10-task DeepSWE codin…
DeepSeek released V4-Flash-Vision-Exp, a multimodal model that outperforms its text-only predecessor on agent benchmarks, scoring 83.9 on Terminal Bench 2.1 (up from 82.7) and 59.3 on DeepSWE (up from…
GLM-5.3 and Claude Fable 5 finish within noise of each other on DeepSWE accuracy, but GLM-5.3 costs a fifth as much per task and wins every multi-attempt metric. Fable 5 leads pass@1 by 0.7 points (69…
GLM-5.3, an open-weight model from Together AI, solves 85.9% of DeepSWE tasks at $6.61 each when cascaded with GPT-5.6 Sol, compared to Sol alone at 72.7% for $8.37, making the cascade 13 points bette…
Unsloth released Dynamic v3.0 GGUFs for Qwen3.8-27B, claiming more than 10% better top-1% accuracy at the same size compared to every other provider. The new quants, which work with llama.cpp and Unsl…
DeepReinforce released Ornith-1.5, an open-source AI model that closes the self-improvement loop by proposing its own training tasks, writing its own scaffolds, and generating its own RL rollouts with…
In a comparison of DeepSeek V4 Pro 0813 and GPT-5.6 Sol on the DeepSWE benchmark, a cascade strategy using DeepSeek first and escalating to GPT-5.6 Sol solves 83.0% of tasks at $3.35 each, while GPT-5…
DeepSeek V4 Pro 0813 solves 82.7% of DeepSWE tasks at $8.28 each when escalated to Claude Fable 5 only on failure, outperforming Fable alone (69.7% at $21.63) by 13 points at 62% lower cost, according…
DeepSeek released its V4-Pro model and an open-source developer tool, DeepSeek Harness, on August 13. The official DeepSWE score for V4-Pro is 62.7, up from 12.8 for the preview, with the footnote ind…
DeepSeek released V4 to general availability, with V4-Pro moving to the 0813 build and V4-Flash to 0731, and ClawBox AI sessions now automatically use the new builds. DeepSeek's reported figures show …
DeepSeek has released DeepSeek-V4-Pro-0813, an advanced AI model with enhanced agent capabilities, now available via its website, mobile app, and API. The model scored 62.7 on the DeepSWE benchmark, u…
Google released Gemini 3.7 Flash on August 13, three weeks after 3.6 Flash, with a pricing strategy that cuts costs by half until the end of 2026. The model shows significant improvements on coding be…
OpenAI's price cut for GPT 5.6 Luna by 80% has made AI-powered analytics answers cost less than half a penny each, with Luna achieving 99.8% accuracy on an agentic SQL benchmark at a 5x lower price th…
Google has introduced Gemini 3.7 Flash, a new AI model that costs half the original price of its predecessor, Gemini 3.6 Flash, at $0.75 per million input tokens and $3.75 per million output tokens th…
Google shipped Gemini 3.7 Flash on August 13, three weeks after Gemini 3.6 Flash, with benchmark gains in agentic coding tasks such as DeepSWE v1.1 jumping from 49.0% to 65.3%. The launch price is $0.…
A new AI model scored 65% on the DeepSWE benchmark, but it is not the model Google promised, according to The New Stack. The article highlights the discrepancy between the announced model and the one …