Easy-Peasy.AI Release Notes

OpenAI's GPT-6 Astra is now available in Marky!

Built for the hardest end-to-end work: complex reasoning, coding, computer use, research, and document creation.

Highlights:

• 1.05M-token context window and 128K output
• Always-on reasoning with effort levels up to a new
  Max
  tier (GPT-6 Astra (Max))
• Strong agentic behaviour: infers intent and biases towards action instead of over-asking
• Internet access, image generation/editing, Python tools, and vision

Try GPT-6 Astra now in Marky chat!

Moonshot AI's Kimi K3 is now available in Marky!

The largest open-weights model ever released — a 2.8-trillion-parameter Mixture-of-Experts model built for long-horizon coding and agentic work.

Highlights

• 1M-token context window
  for huge codebases and documents
• Always-on step-by-step reasoning
• #1 on Arena.AI's Frontend Code Arena; top-3 on GDPval-AA v2 real-world work benchmark, ahead of Claude Opus 4.8
• Internet access, image generation/editing, Python tools, and vision in Marky

Try Kimi K3 now in Marky chat!

OpenAI's GPT-5.6 family is now available in Marky!

What's new:

• Up to 1.05M tokens of context and 128K output

• New maximum reasoning effort for the hardest tasks

• Improved coding, professional work, tool use, and visual design judgment

• Internet access, image generation/editing, Python tools, and vision

• GPT-5.6 Sol

  • Frontier capability for complex professional work, coding, research, and demanding agentic tasks

• GPT-5.6 Terra

  • Strong intelligence with a balanced price for everyday professional workloads

• GPT-5.6 Luna

  • The fastest and most affordable GPT-5.6 model for high-volume work

Try GPT-5.6 now in Marky chat!

Marky can now run tasks automatically on your schedule!

• One-Time or Recurring Runs
  • Schedule tasks once, daily, weekly, or monthly
• Autonomous Agent Execution
  • Marky runs the prompt in the background without needing you online
• Results in Chat
  • Each run is saved back to a Marky conversation so you can review the full output
• Email Notifications
  • Get a notification with a direct link when a scheduled task finishes
• Easy Management
  • View, pause, edit, resume, or delete scheduled tasks from the Scheduled Tasks panel

Try prompts like: "Every morning, research AI news and summarize the top stories."

Four new frontier models are now available in Marky chat!

• GLM 5.2
  • Z.ai's latest flagship with a 1M-token context window and advanced multi-effort coding for long-horizon tasks, powered by the new IndexShare architecture
• Kimi K2.7 Code
  • Coding-focused agentic model built on Kimi K2.6 for long-horizon software engineering workflows
• MiniMax M3
  • Native multimodal MoE model with a 1M context window for text, image, video, and agentic coding tasks
• Qwen 3.7 Plus
  • Alibaba's latest flagship model with strong reasoning and coding capabilities

🧠 Long context · 🌐 Internet access · 🎨 Image generation/editing · 🐍 Python · 👁️ Vision

Try the new models now in Marky chat!

OpenAI's newest frontier model GPT-5.5 is now available!

GPT-5.5 Instant

• Faster, smarter everyday model with native computer use

GPT-5.5 Thinking

• Medium reasoning effort for stronger multi-step problem solving

GPT-5.5 Pro

• Extra-high reasoning effort for maximum performance on complex professional tasks

Key Improvements over GPT-5.4:

• Higher intelligence across reasoning, coding, and agentic benchmarks
• New extra-high (xhigh) reasoning effort for the Pro variant
• 1M token context window with internet, image generation/editing, and Python tools

Try GPT-5.5 now in Marky chat!

DeepSeek V4 Pro is now available — flagship open-source Mixture-of-Experts model for frontier reasoning, advanced coding, and long-context intelligence!

• Unlimited for 1 Month
  • Use DeepSeek V4 Pro at zero credit cost for the next month
• 1M Context Window
  • Handle massive documents, codebases, and datasets in a single prompt
• Frontier Reasoning
  • Intelligence Index of 52 for deep multi-step reasoning and complex agentic workflows
• Advanced Coding
  • Top-tier open-source performance on demanding production workloads
• Hybrid Attention Architecture
  • Dramatically improved long-context efficiency with reduced KV and compute overhead

🧠 1M context · 🌐 Internet access · 🎨 Image generation/editing · 🐍 Python · 💭 Reasoning

Try DeepSeek V4 Pro now in Marky chat!

OpenAI's latest image model

GPT Image 2

is now available in three quality tiers!

• Sharper Detail & Realism
  • Photorealistic output with crisp fabric, hair, and skin textures
• Best-in-Class Text Rendering
  • Accurate typography, logos, and in-image text
• Strong Prompt Adherence
  • Follows complex multi-object prompts with precise composition
• Three Quality Tiers
  • Low (0.5 credits) for drafts, Medium (3 credits) for production, High (11 credits) for maximum fidelity
• Native Image Editing
  • Edit with reference images; preserves subject identity at high fidelity

Try it now in the AI Image Generator!

Anthropic's newest flagship Opus model Claude Opus 5 is now available (July 2026)!

A Step Change Over Opus 4.8

• Stronger deep reasoning, agentic coding, and long-horizon work — at the same price

Built for Complex Coding

• Multi-file features, larger refactors, and end-to-end feature work it actually finishes

Sharper Bug Finding

• High precision and high recall on code review, even at lower reasoning effort

1M Context Window

• Process massive documents, codebases, and datasets in a single prompt (128K output)

Adaptive Thinking

• Selectable effort levels (Low → Max); Low and Medium punch well above their weight on this model

🧠 1M context · 🌐 Internet access · 🎨 Image generation/editing · 🐍 Python · 👁️ Vision

Available in both standard and reasoning variants!

Anthropic's most agentic Sonnet model is now available (June 2026)!

A Drop-In Upgrade to Sonnet 4.6

• The best combination of speed and intelligence, at the same price

Approaches Opus 4.8 Quality

• Closes the gap with Opus-class models on coding, reasoning, and agentic tasks — at a fraction of the cost

Most Agentic Sonnet Yet

• Stronger planning, tool use, and long-horizon autonomous work

Adaptive Thinking

• Thinks by default with selectable effort, or pick the Reasoning variant for the deepest chains of thought

1M Context Window

• Process massive documents, codebases, and datasets in a single prompt (128K output)

🧠 1M context · 🌐 Internet access · 🎨 Image generation/editing · 🐍 Python · 👁️ Vision

Available in both standard and reasoning variants!

Anthropic's most capable widely released model Claude Fable 5 is now available (June 2026)!

Anthropic's Most Capable Model

• Built for the most demanding reasoning and long-horizon agentic work

Always-On Adaptive Thinking

• Thinks by default with selectable effort levels (Low → Max) to balance reasoning depth against latency

1M Context Window

• Process massive documents, codebases, and datasets in a single prompt (128K output)

Built for Agents

• Stronger long-horizon agentic coding, sharper judgment, and reliable tool use

🧠 1M context · 🌐 Internet access · 🎨 Image generation/editing · 🐍 Python · 👁️ Vision

Available in both standard and reasoning variants!

Anthropic's newest flagship model Claude Opus 4.8 is now available (May 2026)!

Anthropic's Most Capable Model

• Reported to surpass GPT-5.5 and Gemini 3.1 Pro across coding, agentic, and reasoning tasks

Long-Horizon Agentic Coding

• Better long-context handling, fewer compactions, and fewer skipped tool calls than Opus 4.7

More Honest

• Clearer, more reliable reporting of its own progress and limitations

1M Context Window

• Process massive documents, codebases, and datasets in a single prompt (128K output)

Adaptive Thinking

• Selectable effort levels (Low → Max), defaulting to High, to balance reasoning depth against latency

🧠 1M context · 🌐 Internet access · 🎨 Image generation/editing · 🐍 Python · 👁️ Vision

Available in both standard and reasoning variants!