PromptsZone Logo

PromptsZone

AI news & analysis

New Models

Original five-row scorecard comparing GPT-6 Cyber with GPT-6 Astra on OpenAI-reported cybersecurity evaluations, with the official OpenAI provider wordmark.

OpenAI releases GPT-6 Cyber to approved security teams, with access still limited

November 5th

GPT-6 Cyber begins a restricted rollout for approved security teams. Its benchmark gains need to be read alongside access requirements and mixed external results.

Read more
New Models
Original explanatory diagram linking the Granite 4.2 8B base to TwIL-8B balanced and verbose variants, with the official webAI provider logo.

webAI adds a balanced TwIL-8B checkpoint for noncommercial research

November 5th

TwIL-8B's balanced checkpoint targets shorter reasoning. Its noncommercial license and mixed benchmark methods are essential context for researchers.

Read more
New Models
Original diagram showing agent text and microphone audio feeding Soniox STT v6 recognition to produce a user transcript with agent playback filtered out.

Soniox releases STT v6 to help voice agents listen while they speak

November 5th

Soniox’s new speech models are active, with real-time text context and echo filtering. Benchmark and output-format caveats matter before migration.

Read more
New Models
Officials seated at microphones during the EKA-1 and Dhīr model launch in New Delhi on November 3, 2026.

IndiaAI unveils EKA-1 and Dhīr, with access and evaluation details still to check

November 5th

IndiaAI’s November 3 model launch pairs Soket’s EKA-1 with ZenteiQ’s compact Dhīr models. Access routes and benchmark evidence need careful reading.

Read more
New Models
Original editorial diagram of Images 2.6 showing Flare and Sunburst, markup and complexity controls, and GIF and SVG output, with the official OpenAI provider wordmark.

OpenAI releases Images 2.6 with faster generation and more editing controls

November 4th

Images 2.6 adds GIFs, API markup and SVG output. The release’s workflow claims and developer-reported safety results call for different kinds of evaluation.

Read more
New Models
Original diagram of 96 file tiles beside the official Kimi family logo, illustrating K3.1 Pro public weights and its custom license.

Kimi K3.1 Pro weights are available, with a custom license for deployment

November 4th

Moonshot’s public model files turn a promised release into a deployment option. The custom license and infrastructure requirements need careful review.

Read more
New Models
Original chart comparing Sarvam Prism and four decision-model peers across three developer-reported tests.

Sarvam Prism launches with open weights for fast, fixed-choice decisions

November 4th

Sarvam Prism scores predefined answers instead of generating text. Compare its developer-reported benchmarks, API access and local hardware requirements.

Read more
New Models
Four-model chart of Tencent-reported finance-task scores: Hy4 3.04, DeepSeek 2.97, Kimi and GLM 2.68. Not an independent overall ranking.

Tencent Hy4 launches through hosted services, with weights still to come

November 4th

Hy4’s access options and a cautious reading of Tencent’s finance-task comparison.

Read more
New Models
LongCat-2.5 benchmark scorecard with Meituan provider logo: Claude Opus 5 leads the selected Test Writing row at 62.2; GLM-5.3 leads SpreadsheetBench 2 at 35.8. Developer-run results, not independent tests.

Meituan releases LongCat-2.5 through its API, with model downloads still pending

November 2nd

LongCat-2.5 reaches Meituan’s API after its September preview. A sourced comparison, current pricing and the distinction between API access and pending downloads.

Read more
New Models
Original scorecard comparing OpenAI-reported Rosalind-6 and rerun Rosalind-5.5 results on MedChemBench v7, GeneBench Pro v13 and LifeSciBench Gold. Higher scores are bold; methods and limitations are labeled.

OpenAI rolls out GPT-6 Rosalind Research with restricted access and new benchmark results

October 30th

OpenAI’s October 30 Rosalind release brings GPT-6 to approved life-sciences workflows, with three benchmark comparisons, higher API prices and restricted access.

Read more
New Models
Original Rime Prism τ-Voice scorecard: Prism 56.6% developer-reported, GPT-Live-1 Sol low 59.3%, GPT-Realtime-2.1 High 45.7%, Gemini 3.8 Live 30.1% from Artificial Analysis. Best displayed score bold; reasoning settings and mixed provenance labeled.

Rime launches Prism, bringing a text checkpoint to enterprise voice AI

October 30th

Rime’s October 28 speech-to-speech release emphasizes inspectable responses and deployment control. Trial access, tuning availability and benchmark limitations need separate checks.

Read more
New Models
Original Particle-1 availability card: API access, text and image inputs with text output, tool calling, WebSocket steering, combined context and default output limits. Documented facts, not benchmark rankings.

Particle launches Particle-1, targeting more efficient coding agents

October 30th

Particle's October 28 model release brings a coding-focused API, long-context input and live steering, with important limits around evaluation, output and billing.

Read more
API New Models
Benchmark scorecard comparing Gemini 4 Carbon, Gemini 4 Argon, GPT-6 Astra, Claude Opus 5.5 and Claude Fable 5.1 across nine tests reported in Google's October 28, 2026 launch table. Independent and mixed-source results are labeled; missing results use dashes.

Google announces Gemini 4 Carbon, with cyber defenders first in line

October 29th

Google’s new frontier model leads an independent benchmark, but its initial rollout is restricted to selected cyber defenders.

Read more
New Models
Original Mistral Large 4 scorecard: independently verified Artificial Analysis Intelligence Index v5.0 score 26 for high reasoning; Mistral-reported Coding Agent Index 48%. Different metrics, checked October 29, 2026.

Mistral releases Large 4 weights, bringing deployment control and licensing trade-offs

October 29th

Mistral's October 29 release makes Large 4 weights downloadable. Coding results, infrastructure requirements and a model-as-a-service license restriction shape the trade-offs.

Read more
New Models
Original scorecard of Giotto-reported Alto and Solo results: AIME 2026 100.0% and 96.6%, GPQA-Diamond 91.9% and 88.3%, HLE 52.0% and 37.5%, Math-500 both 99.80%. Evaluation settings are incompletely specified; not independently replicated by PromptsZone.

Giotto launches Alto and Solo, putting deployment control at the center

October 29th

Giotto's October 27 launch offers European cloud and private-deployment options. Its published benchmark scores and commercial terms need separate scrutiny.

Read more
New Models
2/3
1 2 3

Newsletter Subscription

Receive the latest AI news

Blog logo

PromptsZone

AI news and analysis from PromptsZone: model launches, research, AI tools, and the business and policy developments shaping artificial intelligence.

RSS feed

© PromptsZone, 2026.

Made with Blogmaker

×