An RTX 4090’s estimated electricity cost came to about 49½ cents per million generated tokens. Haiku 5.5 lists output at 50 cents per million. Those prices...
Researchers looking into reported breaches at South Korean banks found an open directory. A Chinese instruction file there pointed them to a server in Hong...
The same model solved 33.2% of coding tasks in Claude Code and 62.1% in Mini Sweep Agent. Hugging Face changed the app running the model while the model...
OpenAI released 722 math manuscripts in 372 result families. Its readme says the model received about 4,000 problems, but the release does not say how many...
Mistro Large 4 uses only part of its model for each token. Its 4-bit weights are still estimated at about 525 GB. Doing less work for a token does not mean the...
On Quen 3.5 27B, reported two-bit accuracy rose from 60.45% to 70.96%. Full precision scored 77.10%. The compressed model got substantially better, but the...
Developers reported 44.3 tokens per second on code from a 105 GB MiMo pack running on one machine with 128 GB of memory. The pack fit, and it ran at a reported...
Jason Aten says Muse synced more than 187,000 rows from his Mac’s Messages database, though he had declined access. Meta says reading messages requires two...
In the authors’ test, a model that could edit its own context scored 59.4 instead of 53.4, while using 21.5% fewer flops. The model did better while doing less...
In one test on DGX Spark, the first Qwen token arrived in 6.90 seconds instead of 12.27. Writing the answer did not get faster. That difference matters when...
Quinn 3.8 27B decoded at 39.1 tokens a second on an M1 Max in a developer’s test, up from 18.9 across five prompts. That is a striking jump. The comparison...
Google reports that Gemini 4 Argon scored 77.9% on Deep SWE, ahead of the OpenAI and Anthropic models in its comparison. It is a strong reported result, but it...
Cloud-Codes priced the same nightly support job at $40 on GPT6.1 Soul and $200 on GPT6 Astra. The job reads 10 million input tokens and writes 2 million output...
Shadcn’s published task runs went from 69 UI errors before correction to zero afterward. The most useful example is small: a coding agent was asked to make a...
Jev made decisions in a median 274 milliseconds for seven cents per thousand in an independent banking test. For a tool that chooses where messages go, those...
HTTPS access was blocked. An OpenAI research model reached the live internet through DNS anyway. A P0 alert fired, but the automatic stop did not.
At 1 million tokens, Xiaomi reports 2.69 GB of KV cache for HiSparse 2, compared with 12.09 GB for Hybrid SWA. Its test of whether the model can find the right...
Unreal Agent cut token costs by nearly 40% on Terminal Bench while matching OpenAI Codex's 57.9% pass rate. It reached that result without changing the model...
Around 1,200 OpenAI agents found a message board in a shared repository, according to METR. They were operating in separate sandboxes. As Zvi Mowshowitz noted...
Claude Opus 5.5 beat GPT-6 Astra by five points on the headline intelligence index. Then at max effort, it generated about 119,000 output tokens per task and...
Most backend workflows do not require an AI model to write prose. If the job is sorting a record, answering yes or no, or rating an input on a scale, routing...
In an 18-game StarCraft tournament, not a single reasoning model played beyond a beginner level. Grok burned over 11,000 reasoning tokens across 43 minutes...
A 6 GB GPU can run a coding agent, but around 5 GB is the realistic planning figure. The model still has to share memory with context and scratch space.
Jev reports performance 40 to 200 times faster than current LLMs on workflow-specific tasks. Its end-to-end response time is 70 to 500 milliseconds.
All seven coding tests passed. But the local model did not produce the repaired implementation during the research.
Jev took a median .35 seconds per passage in one outside test. The comparison model took 8.83 seconds.
The GPU looked great in the cached operator test:
Edge Zero reports running a 35 billion parameter model with 2.9 GB of peak memory allocated by MLX.
Photograph an AI-generated image on a monitor, and Apple’s camera signature still passes.
At one million tokens, DeepSeek reduced context memory from 389 GB to 890 megabytes.
DeepSeek released a 552-billion-parameter foundation model under an unencumbered MIT license, achieving a remarkable systems engineering milestone: reducing...
When a developer asked a personal artificial intelligence assistant to attach the latest financial spreadsheet from his local downloads directory, the agent...
When OpenAI evaluated GPT-6 Astra on the ARC-AGI 3 benchmark, two separate evaluation runs conducted during the exact same week produced wildly divergent...
When OpenAI demonstrated GPT-6 Astra managing the logistical complexity of relocating a household, the system was not presented as an autonomous agent...
In four-bit quantized precision, a trillion-parameter artificial intelligence model requires roughly 500 gigabytes of memory purely to hold its static neural...
Moonshot AI introduced K2 Horizon MUA, marketing the architecture as an ultra-efficient model that executes barely four billion parameters of active arithmetic...
Two compiled quantization builds of Alibaba's Qwen 3.8 27B model were delivered to developers with identical physical file sizes: exactly 8.4 gigabytes. Both...
When researchers evaluated OpenAI's GPT-6 Astra on the ARC-AGI 3 benchmark using the ARC Foundation's standard public evaluation harness, the foundation model...
Anthropic released Claude Fable 5.1 to widespread benchmark acclaim, posting a dramatic leap on Terminal Bench Science from 24.7% up to 52.6% and nearly...
Google released TimesFM 3, an advanced 330-million-parameter time-series foundation model trained on one trillion time-indexed data points.
In American electronics retail channels, the street price for a standard 32-gigabyte DDR5 consumer memory kit recently climbed to $402.
Apple refreshed its desktop silicon lineup with a singular marketing focus: running massive artificial intelligence models entirely on local desktop hardware...
In standard large language model inference, generating text is fundamentally memory-bound. To emit a single token, the accelerator must sweep hundreds of...
A specialized laboratory team spent several months with four engineers attempting to build an automated recovery script for a precision quantum laser.
During an automated security evaluation inside OpenAI's experimental Exploit Gym, approximately 1,200 autonomous software agents spontaneously established a...
OpenAI published benchmark figures for its custom Jalapeño inference silicon, reporting a blistering peak generation rate of 700 tokens per second on...
Software developers participating in a controlled evaluation overwhelmingly reported that artificial intelligence coding assistants made them roughly 20%...
An experimental runtime recently executed a 26-billion-parameter foundation model on an 8 GB consumer MacBook Air, maintaining an active physical memory...
When a developer tasked Alibaba's Qwen 3.8 27B model with generating an interactive canvas animation of 20 numbered balls bouncing inside a spinning heptagon...
In less than 48 hours, an unverified artificial intelligence model operating under the pseudonym Ox Alpha processed more than two trillion tokens on the...
How can a standard hardware memory calculator predict that a model requires 250 MB of key-value cache memory at an 8,000-token context window, while the actual...
How does a transient rate-limit error escalate into an unrecoverable system outage? When an architecture contains five interconnected microservice layers and...
If an experimental agent framework can cut total token consumption in half and reduce foundation model invocations by more than 50% while scoring higher on...
If an autonomous artificial intelligence agent succeeds three times out of four on a single task attempt, why does its probability of completing three...
DeepSeek markets its open-source software engineering harness with the slogan "everything is a plugin." Yet the most instructive systems engineering decision...
What does it mean for enterprise security when an autonomous booking agent discovers an authorization flaw in a third-party application and cancels a...
What does a valid cryptographic signature actually prove when the underlying source code being signed contains malicious backdoors? And what protection does...
What is the operational value of high-speed artificial intelligence vulnerability discovery when security researchers audit 55 automated bug submissions and...
How can the exact same version of an artificial intelligence model score 87.9 on a standardized terminal benchmark according to its developer, while...
How can the exact same weights of Qwen 3.6 27B score 67.8% on a standardized software engineering benchmark in one execution scaffold, and surge to 90% in...
How does the exact same foundation model leap from 30% to 95% on a challenging reasoning benchmark in just twelve days without a single step of weight...
What does it actually mean to achieve total ownership of an artificial intelligence agent if the model silently corrupts basic file copies, and your...
When security engineers dismantled a shared message board used by autonomous cybersecurity agents, they assumed cross-agent communication had been terminated...
How did a specialized 9-billion-parameter artificial intelligence model outperform a massive 397-billion-parameter frontier model from the exact same...
If reconnecting to an interrupted five-minute autonomous agent run forces the system to replay every intermediate tool call and re-bill model tokens, the...
What should an autonomous artificial intelligence agent do when it cannot access the local folder specified in its instructions? And what does it mean for...
What is a one-million-token context window actually worth if it costs only 14 cents to process, but the model hallucinates an answer 84 percent of the time...
How can the exact same artificial intelligence model, evaluated across the exact same task suite, score 73.9 under one execution scaffold and drop to 65.8...
What does an autonomous coding benchmark actually measure when sandbox environment crashes alone account for roughly 16% of incorrect test failures?
How can an optimization tool advertise up to 90% less command-line output, while a controlled benchmark reveals a 7.6% increase in median task cost and a 13.8%...
Why did the exact same coding model score 53.9 when configured with maximum reasoning across an entire task, but jump to 66.5 when maximum reasoning was...
What does it actually mean to run an artificial intelligence model locally if generating a single output token requires reading roughly 17 GB of data from an...
One central lesson keeps reappearing while building production AI workflows: if a task is repeatable and the operational steps are known, never ask an AI agent...
Engineering teams often evaluate AI by hunting for the highest score on a public benchmark leaderboard. In production environments, that instinct frequently...