The Frontier Model Landscape, June 2026 Edition

Last updated: June 11, 2026

The frontier moved three times in the last four months. Anthropic shipped Claude Fable 5 on June 9, the first publicly available Mythos-class model. OpenAI consolidated around GPT-5.5 and the GPT-5.4 family. Google's Gemini 3.1 Pro is still technically a preview but is deployed everywhere. And DeepSeek V4 reset the open-weights price floor with a 1M-token context window at pennies per million tokens.

This is a state-of-play directory, not a leaderboard. Every price below was read from the vendor's official pricing page on June 11, 2026. Each entry gets one honest paragraph plus a best-for call, with deeper head-to-head comparisons linked throughout.

How This Directory Is Ordered

Models are grouped by tier - maximum capability, frontier workhorse, and budget frontier - rather than ranked on a single number, because benchmark figures come from different reporters using different harnesses, and a unified ranking would be false precision.

One pricing caveat up front: Anthropic's models from Opus 4.7 onward (including Fable 5) use a new tokenizer that can produce up to 35% more tokens for the same text, per Anthropic's pricing docs, so sticker prices understate the real cost jump versus older models.

Maximum Capability Tier

1. Claude Fable 5 (Anthropic)

Fable 5 went GA on June 9, 2026 across the Claude API, Bedrock, Vertex AI, and Microsoft Foundry, per Anthropic's models overview. It is the first public model from the restricted Mythos line: 1M-token context, 128K max output, January 2026 knowledge cutoff, always-on adaptive thinking, and safety classifiers that can refuse cybersecurity and biology requests, with an optional automatic fallback to Opus 4.8. Simon Willison's release-day verdict captures the tradeoff: he calls it "a beast" that is "slow, expensive" but exceptionally capable, with deeper knowledge recall than Opus 4.8. At $10 input / $50 output per MTok it costs exactly double Opus 4.8, and it is included in Claude subscription plans only through June 22. The honest read: this is an async heavy-lift tool, not a default. Our Fable 5 vs Opus 4.8 decision guide covers when the premium pays off.

Best for: long-horizon agentic coding, multi-file migrations, underspecified projects, and research-grade analysis where a failed run costs more than the tokens.

2. GPT-5.5 and GPT-5.5-pro (OpenAI)

GPT-5.5 sits at the top of OpenAI's current pricing page at $5 input / $30 output per MTok, with cached input at $0.50 and a 50% batch discount. The pro variant runs $30 / $180 - the most expensive mainstream API price in this directory. On the evidence we could verify, GPT-5.5 is the strongest non-Anthropic option for agentic terminal work: DataCamp's DeepSeek V4 analysis reports it at 58.6% on SWE-bench Pro and 82.7% on Terminal-Bench 2.0, ahead of every open-weight model. The honest read: GPT-5.5 is priced like a near-peer to Opus 4.8 ($5 input on both, $30 vs $25 output) and the choice is genuinely close - see our GPT-5.5 vs Claude Opus 4.8 comparison and the Fable 5 vs GPT-5.5 benchmark breakdown.

Best for: teams on OpenAI tooling that want frontier agentic capability without Fable-tier pricing, and terminal-heavy agent harnesses.

3. Claude Mythos 5 (restricted)

Listed for completeness, not for selection. Mythos 5 shares Fable 5's $10 / $50 pricing, 1M context, and 128K output, and is available only to approved Project Glasswing customers, per Anthropic's models overview. It ships without Fable 5's stricter safety classifiers, per Simon Willison's launch-day notes. Unless you are a cyberdefense or critical-infrastructure organization with an Anthropic account team, this is not a model you can buy. Our explainer on what Claude Mythos 5 is and who it is for covers the access program.

Best for: approved Glasswing organizations only. Everyone else uses Fable 5.

Frontier Workhorse Tier

4. Claude Opus 4.8 (Anthropic)

At $5 / $25 per MTok with a 1M context window, 128K output, and a January 2026 knowledge cutoff, Opus 4.8 is still the recommended starting point in Anthropic's own models overview even after Fable 5 shipped. The full 1M window bills at standard rates - no long-context premium, per the pricing page - and a fast mode research preview offers quicker output at $10 / $50. The honest read: for interactive coding, code review, and most production agent work, Opus 4.8 is the best capability-per-dollar in the Anthropic lineup.

Best for: the default frontier coding model - interactive development, code review, production agents with cost ceilings.

5. GPT-5.4 family (OpenAI)

Released March 5, 2026, GPT-5.4 absorbed GPT-5.3-Codex's coding stack into the mainline model, per nxcode's GPT-5.4 guide, which reports 57.7% on SWE-bench Pro and 75% on OSWorld - above the 72.4% human baseline on desktop automation. OpenAI's pricing page lists the family at $2.50 / $15 (standard), $0.75 / $4.50 (mini), and $0.20 / $1.25 (nano), with cached input at one tenth of base. One catch: 272K standard context, 1M via API, with input pricing doubling to $5.00 above 272K per nxcode. The honest read: GPT-5.4 is the value pick of the closed-model field - Sonnet-class pricing with near-flagship agentic scores. See our GPT-5.4 vs Gemini 3.1 Pro vs DeepSeek V4 three-way.

Best for: cost-conscious production workloads, computer-use agents, and tiered routing where mini and nano absorb the easy traffic.

6. Gemini 3.1 Pro (Google)

Google released Gemini 3.1 Pro in preview on February 19, 2026, headlined by a verified 77.1% on ARC-AGI-2 - more than double Gemini 3 Pro's score. Nearly four months later it is still labeled preview on the Gemini API pricing page, at $2 / $12 per MTok up to 200K tokens and $4 / $18 beyond, though it is deployed across AI Studio, Gemini CLI, Antigravity, Android Studio, Vertex AI, and the Gemini app. The honest read: the abstract-reasoning numbers are the best in this directory and the price is the lowest of any closed Pro-tier model under 200K context, but the long-context surcharge and the lingering preview label make it harder to commit to for regulated production use. A Gemini 3.5 Pro does not appear on the pricing page yet, so it is not in this directory. Our Claude Fable 5 vs Gemini 3.1 Pro comparison has the head-to-head.

Best for: reasoning-heavy workloads under 200K context, teams inside the Google Cloud ecosystem, and multimodal pipelines.

Budget Frontier Tier

7. DeepSeek V4 (open weights)

DeepSeek released the V4 preview on April 24, 2026: V4-Pro at 1.6T total / 49B active parameters and V4-Flash at 284B / 13B, both MoE models with sparse attention, a 1M-token context window, and weights downloadable from Hugging Face, per the official announcement. The API speaks both OpenAI and Anthropic formats. Official API pricing is the story: V4-Pro at $0.435 input / $0.87 output and V4-Flash at $0.14 / $0.28, with cache hits under a cent. Some third-party comparisons cite higher V4-Pro figures; the official page as fetched on June 11 shows the numbers above. DataCamp's benchmark roundup puts V4-Pro at 55.4% on SWE-bench Pro versus GPT-5.5's 58.6%, confirms the MIT license, and positions DeepSeek as trailing closed state-of-the-art by three to six months at a fraction of the price. That framing is correct: you give up the last few benchmark points and accept a China-hosted API or self-hosting, and you get frontier-adjacent capability at one to two orders of magnitude lower cost. The legacy deepseek-chat and deepseek-reasoner names retire July 24, 2026 - our deepseek-chat to V4 migration guide covers the switch, and the DeepSeek V4 developer guide covers setup.

Best for: high-volume agent loops, cost-sensitive products, self-hosting and fine-tuning, and any workload where 90% of frontier quality at 3% of frontier price is the right trade.

Verified Pricing Table

All prices in USD per million tokens, read from official vendor pricing pages on June 11, 2026. Sources: Anthropic, OpenAI, Google, DeepSeek.

Anthropic, OpenAI, and Google each offer a 50% batch discount; DeepSeek's pricing page does not list one. The spread is the headline: DeepSeek-V4-Flash output is roughly 178x cheaper than Fable 5 output. Full cost modeling is in our frontier model API pricing breakdown.

How to Actually Choose

Two questions settle most routing decisions. Is the task long-horizon and autonomous? If yes and budget allows, Fable 5; if budget matters, Opus 4.8 or GPT-5.5. Is the workload high-volume and tolerant of a small quality gap? DeepSeek V4-Flash or GPT-5.4-nano will cut your bill by 10-100x. If you are moving an existing Claude workload up a tier, start with our Fable 5 migration guide - the thinking-parameter and tokenizer changes break naive drop-in swaps.

FAQ

What is the most capable AI model in June 2026?

Claude Fable 5, by most early accounts including Simon Willison's release-day testing. But "most capable" is task-dependent: GPT-5.5 leads the verified Terminal-Bench numbers among non-Anthropic models, and Gemini 3.1 Pro holds the strongest published ARC-AGI-2 score at 77.1%.

Is DeepSeek V4 really a frontier model?

Close enough to matter. DataCamp's analysis puts V4-Pro within about three points of GPT-5.5 on SWE-bench Pro (55.4% vs 58.6%) at $0.435 / $0.87 per MTok on the official API, with MIT-licensed weights downloadable. The honest framing: three to six months behind closed state-of-the-art.

Why is Gemini 3.1 Pro still in preview?

Google has not said. It launched in preview on February 19, 2026 with GA "coming soon," and the pricing page still labels it gemini-3.1-pro-preview as of June 11. Teams with strict production-SLA requirements should factor that in.

Should I pay 2x for Fable 5 over Opus 4.8?

Only for long-horizon agentic work where quality reduces total attempts. Anthropic's own docs still recommend Opus 4.8 as the starting point for complex tasks. The math is covered in our Fable 5 vs Opus 4.8 guide and cost-per-task analysis.

What is Claude Mythos 5 and can I use it?

Mythos 5 is Fable 5 without the safety classifiers, offered only to approved organizations through Anthropic's Project Glasswing at the same $10 / $50 pricing. There is no self-serve access. For practical purposes, Fable 5 is the Mythos-class model you can actually buy.

Sources

All accessed June 11, 2026.

• Anthropic: Claude API pricing - Claude model pricing, caching, batch, long-context policy, tokenizer note
• Anthropic: Models overview - Fable 5 GA date, context windows, cutoffs, Mythos 5 availability
• OpenAI: API pricing - GPT-5.5 and GPT-5.4 family pricing and batch discount
• Google: Gemini API pricing - Gemini 3.1 Pro Preview tiered pricing and batch rates
• DeepSeek: API pricing - V4-Pro and V4-Flash pricing, context, deprecation date
• DeepSeek: V4 preview announcement - release date, parameters, architecture, weights, API compatibility
• Google: Gemini 3.1 Pro announcement - release date, ARC-AGI-2 score, preview status
• Simon Willison: Initial impressions of Claude Fable 5 - launch details and qualitative assessment
• nxcode: GPT-5.4 complete guide - release date, variants, context structure, benchmarks
• DataCamp: DeepSeek V4 analysis - benchmarks vs GPT-5.5, MIT license, positioning