Updates from the Kyma team.

Launches, guides, and lessons from running every open model behind one API.

Best Long-Context LLMs in 2026: 200K–1M Token Models Compared
Latest

Best Long-Context LLMs in 2026: 200K–1M Token Models Compared

An honest ranking of the best long-context models in 2026 — 200K to 1M token windows for large codebases, RAG, and document analysis. Real specs and prices.

·4 min read

More posts

20

Kyma now runs on Cloudflare's edge

What happened during today's 4-hour outage, why we moved Kyma's primary backend to Cloudflare Workers the same evening, and what it means for you.

·4 min read
GPT Image 2 is live on Kyma

GPT Image 2 is live on Kyma

OpenAI's flagship image model — near-perfect text-in-image, multilingual typography, photoreal output. One SKU, three quality tiers (low / medium / high). $0.014 — $0.297 per 1024×1024.

·4 min read
Kyma now has a voice

Kyma now has a voice

Text-to-speech, prompt-driven music, and sound effects — three new audio capabilities behind your existing Kyma key. Build podcasts, voice agents, and full video soundtracks without a separate billing account.

·3 min read
Image catalog refresh: Recraft V4, FLUX.2 Pro, and native SVG

Image catalog refresh: Recraft V4, FLUX.2 Pro, and native SVG

Five new image SKUs on Kyma: top-of-Arena Recraft V4, BFL's 32B FLUX.2 Pro with multi-reference blending, and the only image gen API in the world that ships actual SVG files.

·4 min read
Kyma now hears

Kyma now hears

Two new audio endpoints turn any agent into one that can listen. Pull the words out of a podcast, read the mood of a clip, watch any social video for under a cent. One key, every modality.

·4 min read
Video generation is live on Kyma

Video generation is live on Kyma

Five video models, one async endpoint, per-second pricing from $0.0945 to $0.410. Kling 3 Pro for cinematic hero clips, Kling 3 Pro Audio for native sound, Seedance 2 Pro for multi-shot action, Seedance 2 Fast for social. T2V or I2V from the same call.

·5 min read
DeepSeek V4 is live on Kyma

DeepSeek V4 is live on Kyma

DeepSeek's V4 family - V4-Pro (1.6T MoE flagship) and V4-Flash (284B value tier) - now live on Kyma. 1M context, native reasoning, MIT license. From $0.19 / $0.38 per 1M.

·5 min read
Image generation is live on Kyma

Image generation is live on Kyma

Four image models, one API key, $0.054 to $0.108 per image. FLUX 1.1 Ultra for photoreal hero shots, Ideogram V3 for typography that actually renders, Recraft V3 for vector and brand, FLUX Kontext Pro for image edits.

·4 min read
Kimi K2.6 is live on Kyma

Kimi K2.6 is live on Kyma

Moonshot's newest agentic flagship (April 2026) — 262K context, native vision, 300-agent swarm. Now live at $1.28 / $5.40 per 1M tokens, or via the agent alias.

·3 min read
How to Build an AI Agent Backend with Open-Source Models

How to Build an AI Agent Backend with Open-Source Models

Build a production-ready AI agent with tool calling, multi-step reasoning, and conversation loops — using open-source models via Kyma API.

·3 min read
Best Open-Source Models for AI Coding Agents in 2026

Best Open-Source Models for AI Coding Agents in 2026

A practical guide to choosing the right open-source model for your coding agent, IDE, or CLI tool.

·2 min read
Build a Chatbot with Next.js and Kyma API in 5 Minutes

Build a Chatbot with Next.js and Kyma API in 5 Minutes

Ship a streaming chatbot with model switching, dark theme, and zero config — using Next.js, Vercel AI SDK, and Kyma API.

·2 min read
Open Source LLMs, One API — Start Free with Kyma

Open Source LLMs, One API — Start Free with Kyma

Get $0.50 free credits and access 60+ models — text, image, video, and voice — through a single OpenAI-compatible endpoint.

·3 min read
Access Kyma Models from Claude Code with Kyma's MCP Server

Access Kyma Models from Claude Code with Kyma's MCP Server

Set up Kyma's MCP server to use Kyma models directly from Claude Code, Cursor, or any MCP-compatible tool.

·2 min read
OpenAI-Compatible APIs: What They Are and Why They Matter

OpenAI-Compatible APIs: What They Are and Why They Matter

Switch from OpenAI to open-source models without changing your code. Just update two values.

·2 min read