June 2026 is the most competitive frontier-model landscape yet. Anthropic now ships three distinct Claude tiers — Claude Fable 5 (generally available), Claude Mythos 5 (limited Glasswing release without safety classifiers), and the still-widely-deployed Claude Opus 4.8. OpenAI's ChatGPT 5.6, Google's Gemini 3.5, Zhipu's GLM 5.2, xAI's Grok 4, and DeepSeek V4 round out the field.
Claude Fable 5 vs Claude Mythos 5
Both share identical specs: 1M-token context, up to 128K output tokens, adaptive thinking, and the same API pricing ($10/$50 per million tokens). The difference is policy: Fable 5 includes safety classifiers that can return stop_reason: refusals — you need fallback handling in production. Mythos 5 removes those classifiers for teams that need unconstrained completions, but is only available through Anthropic's Project Glasswing limited release. If you lack Mythos access, Fable 5 is the upgrade path from Opus 4.8.
Claude Opus 4.8 — still relevant?
Yes for teams mid-migration. Opus 4.8 remains excellent for coding and long-form analysis with 200K+ context and mature SDK patterns. Anthropic recommends migrating agentic pipelines to Fable 5 for 1M context, task budgets, and programmatic tool calling — but Opus 4.8 is stable, well-understood, and still powers many Claude Code deployments.
ChatGPT 5.6 (OpenAI)
OpenAI's consumer and API flagship balances reasoning depth with the richest plugin, Custom GPT, and developer ecosystem. Strong multimodal (voice, vision, image gen), dependable enterprise compliance, and the default choice for teams already on Azure/OpenAI. Slightly shorter context than Gemini or Claude Fable, but unmatched integration breadth.
Gemini 3.5 (Google)
Google's answer for users living in Workspace, Android, and Search. Up to 2M-token context handles full codebases and long video. Native multimodal across text, image, audio, and video in one pass. Best pick for research-heavy workflows and Google-centric enterprises.
GLM 5.2 (Zhipu AI)
The standout bilingual model for Chinese–English teams. Competitive API pricing, solid agentic tool use, and growing open-weight releases make GLM 5.2 the value frontier pick — especially for APAC deployments and hybrid stacks alongside Western models.
Grok 4 & DeepSeek V4
Grok 4 differentiates with live X/Twitter data and a witty, news-first personality — ideal for social and media workflows. DeepSeek V4 continues to disrupt on price: open weights, strong reasoning, and API costs a fraction of Claude or GPT — the default for cost-sensitive engineering teams.
How to choose
Software & agents: Claude Fable 5 or ChatGPT 5.6. Google shops: Gemini 3.5. Bilingual CN/EN: GLM 5.2. Budget / self-host: DeepSeek V4. Real-time social: Grok 4. No-refusal Anthropic tier: Claude Mythos 5 (if you have Glasswing access). Most professionals run a primary + a budget fallback.
See also: ChatGPT vs Claude, Claude vs Gemini, and DeepSeek vs ChatGPT.