ONLINEAGENT_OPS 2026.Q3 HOME ARTICLES CRAFT RECORD BLOG MAP HUBS FAQ SEARCH
HOMEARTICLESASTRA, OPUS, FLASH: WHAT AI MODEL NAMES
ARTICLES · EXPLAINER

Astra, Opus, Flash: What AI Model Names Actually Mean

Four vendors, four naming systems, none of which tells you the price. What each scheme encodes, where a higher number misleads, and the rule that survives.

READ4 min
WORDS728
SECTIONS6
TYPEEXPLAINER
CHECKED16 SEP 26

Astra, Sol, Terra, Luna. Fable, Opus, Sonnet, Haiku. 4.6, 3.8, 3.1. Four vendors, four naming systems, none of which tells you what the model does. Here is what each scheme actually encodes, and the one rule that survives all of them.

TL;DR — THE SHORT VERSION
  • OpenAI uses celestial names for tiers within a generation, and a number for the generation.
  • Anthropic uses literary names for size, and a number for the version.
  • Google and xAI use numbers only, plus a word for the speed tier.
  • A higher number is not always a better model — or a newer one.
  • The name never tells you the price. Look it up; the spread is a hundredfold.
◈ PRICES ON THIS PAGE

Every price here was correct when this site last checked it, on 16 Sep 2026. Prices change often and this site no longer updates them, so check the vendor’s own page before you rely on one: xAI pricing.

The four schemes

Current lineups as each vendor lists them, read at source 16 Sep 2026; the Anthropic row 23 Sep 2026
VendorPatternCurrent names
OpenAIgeneration number + celestial nameGPT-6 Astra, GPT-5.6 Sol, Terra, Luna
Anthropicliterary name + version numberFable 5.1, Opus 5.5, Sonnet 5, Haiku 4.5
Googlenumber + tier wordGemini 3.8 Flash, 3.1 Pro, Flash-Lite
xAInumber onlyGrok 4.7, 4.6, 4.5, 4.3

OpenAI, Models · Anthropic, Models overview · Google, Gemini Developer API pricing (read in a browser) · xAI, Pricing. All read 16 Sep 2026.

Anthropic, Models overview, read at source 23 Sep 2026, which lists “Claude Fable 5.1”, “Claude Opus 5.5”, “Claude Sonnet 5” and “Claude Haiku 4.5”. First-hand: until 23 Sep 2026 the Anthropic row read “Fable 5.1, Opus 5, Sonnet 5, Haiku 4.5”.

OpenAI: the number is the generation, the name is the tier

GPT-6 Astra sits above GPT-5.6 Sol, and within 5.6 the order runs Sol, Terra, Luna. OpenAI’s own descriptions carry the hierarchy: Astra is “our most capable model, built for the hardest end-to-end work”, Sol is the “flagship model for complex professional work”, Terra “balances intelligence and cost”, and Luna is “optimized for cost-sensitive workloads”.

The catch is that the tiers are not comparable across generations. A 5.6-series name tells you where it sits inside 5.6, not how it compares to a 6-series model.

Anthropic: the name is the size

Haiku, Sonnet, Opus — short form to long form, cheapest to dearest. The version number moves independently, which is why Haiku 4.5 sits beside Opus 5.5 in the same current lineup. Anthropic calls Opus 5.5, released on 22 Sep 2026, “the first model in our new Claude 5.5 family”, with Sonnet 5.5 and Haiku 5.5 to follow; until they do, the four current models carry four different numbers. What changed from Opus 5 is on Claude Opus 5.5 vs Opus 5.Anthropic, Introducing Claude Opus 5.5, 22 Sep 2026, read at source 23 Sep 2026: “We’re introducing Claude Opus 5.5, the first model in our new Claude 5.5 family.” and “Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks”.First-hand: until 23 Sep 2026 the first sentence here named Opus 5.

Fable 5.1 breaks the poetry pattern and sits above Opus. Anthropic describes it as being for “demanding reasoning and long-horizon agentic work”, and tells you when to reach for it: “when your evals on Claude Opus 5.5 at higher effort still fall short.”Anthropic, Models overview, read at source 23 Sep 2026: “Use Claude Fable 5.1 for demanding reasoning and long-horizon agentic work, or when your evals on Claude Opus 5.5 at higher effort still fall short.”First-hand: until 23 Sep 2026 this quote named Claude Opus 5, as the documentation did when read on 16 Sep 2026.

WHAT THE NAME ENCODES · TWO LADDERS
The OpenAI and Anthropic lineups as listed 16 Sep 2026, the Anthropic side updated 23 Sep 2026, highest tier at the top. The name carries the tier; the number tracks generation or version, which is why Opus is at 5.5, Sonnet at 5 and Haiku at 4.5.
OPENAIANTHROPICNAME = TIERNAME = SIZEAstraGPT-6SolGPT-5.6TerraGPT-5.6LunaGPT-5.6Fable5.1Opus5.5Sonnet5Haiku4.5NUMBER = GENERATIONNUMBER = VERSION
Reasoning — summarises this page’s table of the four schemes and its OpenAI and Anthropic sections, page checked 16 Sep 2026, Anthropic lineup updated 23 Sep 2026.
TAKEAWAY

Of these four schemes, Anthropic’s is the one where the name alone tells you the size tier. Everywhere else you have to look the model up.

Where a higher number misleads

Three real examples from the current lineups:

  • Grok 4.3 is cheaper and has a bigger context window than Grok 4.6 (and the newer 4.7, priced the same as 4.6) — 1M against 500K, at $1.25 against $2.00.xAI, Pricing, read at source 16 Sep 2026: grok-4.3 context “1M” at “$1.25”, grok-4.6 context “500k” at “$2.00”.
  • Claude Haiku 4.5 and Claude Sonnet 5 are both current, but Haiku’s reliable knowledge cutoff is February 2025 against Sonnet’s January 2026 — eleven months apart inside the same lineup.Anthropic, Models overview, read at source 16 Sep 2026: reliable knowledge cutoff “Feb 2025” for Haiku 4.5 and “Jan 2026” for Sonnet 5.
  • Gemini 3.8 Flash is cheaper than Gemini 3.1 Pro, because Flash and Pro are different tiers, not different vintages.

So the number orders a vendor’s own line, and nothing else. It does not order across lines, across vendors, or against price.

The one rule that survives

Look up the price, the context window and the retirement date. Never infer any of the three from the name.

All three are published, all three are checkable in under a minute, and all three are what actually decide whether a model fits. The comparison tables are on LLM API prices, and the dates on when AI models retire.

The wider pattern of products being renamed out from under people is on what got renamed in 2026.

What this page could not verify

  • Why any vendor chose its scheme. None publishes a rationale.
  • Whether the schemes are stable. OpenAI has changed naming conventions repeatedly; nothing says this one lasts.
  • Relative quality. This page is about names, not capability.
SOURCES

OpenAI, Models · Anthropic, Models overview · Google, Gemini Developer API pricing · xAI, Pricing. All read at source on 16 September 2026.

Naming schemes change as often as the models do.

ABOUTMETHODVERIFYPRIVACYCONTACTINDEXAI PROMPT GENEER · EVERY ARTICLE CARRIES ITS OWN CHECKED DATE