Skip to content
TaskPriced
Anthropic · Claude 5.2OpenAI · GPT-6.2cross-vendor· verified 2026-11-18

Claude Fable 5.2 vs GPT-6.2 Astra

Both list at $8/$40 per MTok — why does Claude Fable 5.2 cost 6.6x more per task than GPT-6.2 Astra?

01 // VERDICT

The short answer

Headline numbers
At a glanceClaude Fable 5.2GPT-6.2 Astra
List priceinput / output per 1M tokens$8 / $40$8 / $40
Benchmark task costArtificial Analysis, per Intelligence-Index task$26.54$4.00 (better)
Intelligence IndexAA v5.0.1, higher is better65 (better)62

Highlighted values mark the lower rate or stronger measured score in the same documented run.

EXECUTIVE SUMMARY

On paper, Claude Fable 5.2 and GPT-6.2 Astra carry the exact same headline API price ($8.00 input / $40.00 output per 1M tokens). In independent Artificial Analysis Index v5.0.1 evaluation, however, Fable 5.2 (Max) costs $26.54 per task (41.1 min) versus GPT-6.2 Astra (Max) at $4.00 per task (11.5 min) — a 6.6x realized cost difference driven by Fable's adaptive thinking token volume. Fable 5.2 takes #1 overall on Intelligence Index (65 vs 62) and wins decisively on factual reliability (AA-Omniscience 67 vs 50), HLE-Diamond (80.2% vs 72.2%), CritPt v2 (76.5% vs 71.5%), and AA-Briefcase (1785 vs 1644), plus stays flat $8/$40 across 1M context with 2.5% cache reads ($0.20 vs Astra's $0.40). GPT-6.2 Astra wins Terminal-Bench 4.0 (64.1% vs 59.6%), Terminal-Bench-Science 0.1 (78.1% vs 73.3%), AutomationBench-AA (77.2% vs 76.4%), and GDP.pdf (29.8% vs 24.2%) — and matches Fable High/Xhigh's Index 62 score at $2.52-$4.00 instead of $9.08-$14.83.

RECOMMENDED WHEN

Choose Claude Fable 5.2

Maximum single-shot intelligence (AA Index 65), factual accuracy/low hallucination (AA-Omniscience 67 vs 50), massive prompts (>272K where Fable stays flat $8/$40 vs Astra's $16/$60 cliff), or >90% prompt-cache hit rates ($0.20 vs $0.40).

RECOMMENDED WHEN

Choose GPT-6.2 Astra

Cost-controlled frontier reasoning: equal Index 62 at $2.52-$4.00/task vs Fable's $9.08-$14.83/task, faster end-to-end completion (11.5 vs 41.1 min max), Terminal-Bench 4.0 (64.1% vs 59.6%), AutomationBench-AA (77.2% vs 76.4%), or Ultrafast/background discounts.

TIE-BREAKER RULE

Either — decide by evidence you generate

Tune effort before writing checks: cap Fable 5.2 at medium (Index 56, $3.80/task) or high ($9.08/task) if deploying Anthropic, vs Astra xhigh (Index 62, $2.52/task).

02 // KEY FACTS

Key facts

  • Identical $8.00/$40.00 list rates diverge 6.6x on real Index tasks in AA v5.0.1: Claude Fable 5.2 (Max) costs $26.54/task (41.1 min) vs GPT-6.2 Astra (Max) $4.00/task (11.5 min).
  • Effort-for-effort at Index score 62: GPT-6.2 Astra (Xhigh) costs $2.52/task (6.5 min) vs Claude Fable 5.2 (High) $9.08/task (14.9 min) or (Xhigh) $14.83/task (24.0 min).
  • Context pricing favors Fable above 272K: Fable 5.2 stays flat $8/$40 up to 1,000,000 tokens, whereas Astra jumps to $16/$60 above 272,000 input tokens.
  • Cached reads favor Fable in percentage terms ($0.20 = 2.5% short/long vs Astra $0.40 short / $0.80 long = 5%), while both charge $10.00 short standard cache writes (5m Anthropic / 30m OpenAI).
  • Serving tier asymmetry: Astra adds Ultrafast (3x promo through 2027-02-28) and beta -30%/-40% background windows; Fable supports Standard, Batch (-50%), Fast (2x = $16/$80) at ~111 vs 60 tok/s output speed.
  • Benchmark split in AA v5.0.1: Fable leads Index (65 vs 62), AA-Omniscience (67 vs 50), AA-Briefcase (1785 vs 1644), HLE-Diamond (80.2% vs 72.2%), FrontierCode-AA 2.0 (48.8% vs 45.4%); Astra leads AutomationBench-AA (77.2% vs 76.4%), Terminal-Bench 4.0 (64.1% vs 59.6%), Terminal-Bench-Science (78.1% vs 73.3%), and GDP.pdf (29.8% vs 24.2%).

03 // TIER PRICING

Official pricing, every tier

USD per 1M tokens from vendor pricing pages. 'Not published' means exactly that — we don't extrapolate.

Official API pricing per million tokens, standard context band
Service tierClaude Fable 5.2GPT-6.2 Astra
Standarddefault on-demand serving

in $8 · cached $0.20 · out $40

cache write $10

in $8 · cached $0.40 · out $40

cache write $10 · >272K in/req

Batchasync jobs, 50% off

in $4 · cached $0.10 · out $20

cache write $5 · derived

in $4 · cached $0.20 · out $20

cache write $5 · >272K in/req

Fast2× on supported OpenAI and Anthropic models

in $16 · cached $0.40 · out $80

cache write $20 · derived

in $16 · cached $0.80 · out $80

cache write $20 · >272K in/req

UltrafastOpenAI accelerated servingNot published

in $24 · cached $1.2 · out $120

cache write $30 · >272K in/req

Long-context pricing per million tokens
Long-context bandClaude Fable 5.2GPT-6.2 Astra
Long-context ratesapplies per request above the model threshold
Flat pricing

same rates at any input size (up to context window)

in $16 · cached $0.80 · out $60

cache write $20 · >272K in/req

USD per 1M tokens. Every input token is billed in exactly one bucket — uncached, cached read, or cache write; a write replaces the input rate for those tokens (official rule), so write rates are per tier as listed. OpenAI long context = input > 272,000 tokens per single request. · OpenAI pricing (verified 2026-11-18) · Anthropic rows bill the 5-minute cache-write TTL ($5/MTok on Opus 5.5, $10/MTok on Fable 5.2); 1-hour writes ($8 and $16) are also published, and cache multipliers stack with Batch and Fast-mode pricing. · Anthropic pricing (verified 2026-11-19)

Tier ≠ quality change

Batch/Flex/Fast/Ultrafast change serving speed and price, not the underlying model. Flex and Batch are half price; Fast is 2x; Ultrafast is 6x (GPT-6 / 6.1) or 3x intro promo through 2027-02-28 (GPT-6.2; 6x starting 2027-03-01). On OpenAI these multipliers apply per tier as listed (and OpenAI also offers 2h -30% / 8h -40% beta background windows on GPT-6.2); Anthropic publishes Batch (-50%) and Fast (2x) for Claude Opus 5.5 and Claude Fable 5.2, and both stack with prompt-caching prices.

04 // COST SIMULATOR

Run this comparison on your own numbers

Preloaded with a workload typical for this comparison. Change anything — results recompute locally.

Models (pick up to 4)

Estimated monthly ledger

browser-local

Token-based workload estimate only; it does not measure quality or cost per successful task.

Estimated monthly cost per model
ModelPer requestMonthly
Claude Fable 5.2cheapest$0.375$5,625.00
GPT-6.2 Astra$0.38$5,700.00

Monthly spread between the cheapest and priciest selected model: $75.00.

Breakdown: Claude Fable 5.2 · standard tier · standard context
  • Fresh input (uncached): 10,000 tok × $8/M$0.08
  • Cached input (reads): 25,000 tok × $0.2/M$0.005
  • Cache writes: 5,000 tok × $10/M$0.05
  • Output (incl. reasoning tokens): 6,000 tok × $40/M$0.24
  • Per request$0.375
  • 5,000 tokens billed at the cache-write rate instead of the input rate (mutually exclusive buckets — no double charge).

Rates: official pricing · verified 2026-11-18

  • Per-token arithmetic only: it does not model quality differences or cross-vendor output-token-count differences for the same task.
  • Reasoning tokens are billed as output tokens at the chosen model's output rate.
  • Long-context band is decided per request by total input tokens vs the provider threshold, not by monthly volume.
  • Cache accounting uses the official mutually-exclusive buckets: each input token is billed once, as uncached input, cached read, OR cache write — writes replace the input rate for those tokens, never add to it.
  • Claude rows bill the 5-minute-TTL cache write; the published 1-hour TTL ($8/MTok standard) is not modeled (no TTL input).
  • Rows flagged 'derived' are computed from a stated official rule (e.g. Anthropic batch -50%), not read verbatim from a price table.
  • Model and tier combinations without a published rate appear as Unsupported.

05 // BENCHMARK LOG

What the evidence actually says

Independent and vendor-run evals, each with publisher, runner, and harness. Numbers from different runs are never merged.

Benchmark evidence log with harness attribution
ModelScoreEffortHarnessEvidence
AA Intelligence Indexv5.0.1
GPT-6.2 Astragpt-6.2-astra62 index-pointsconfig 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-265 index-pointsconfig 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
AA cost per Intelligence-Index taskv5.0.1
GPT-6.2 Astragpt-6.2-astra4 usd-per-taskconfig 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-226.54 usd-per-taskconfig 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
AutomationBench-AAAA v5.0.1
GPT-6.2 Astragpt-6.2-astra77.2 %config 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-276.4 %config 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Terminal-Bench4.0 (AA v5.0.1)
GPT-6.2 Astragpt-6.2-astra64.1 %config 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-259.6 %config 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
FrontierCode-AA2.0 (AA v5.0.1)
GPT-6.2 Astragpt-6.2-astra45.4 %config 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-248.8 %config 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
AA-Briefcasev1.1 (AA v5.0.1)
GPT-6.2 Astragpt-6.2-astra1644 eloconfig 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-21785 eloconfig 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
AA-Omnisciencev1.1 (AA v5.0.1)
GPT-6.2 Astragpt-6.2-astra50 index-pointsconfig 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-267 index-pointsconfig 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
HLE-DiamondAA v5.0.1
GPT-6.2 Astragpt-6.2-astra72.2 %config 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-280.2 %config 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
CritPt v2 (Hard Subset)AA v5.0.1
GPT-6.2 Astragpt-6.2-astra71.5 %config 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-276.5 %config 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Terminal-Bench-Science0.1 (AA v5.0.1)
GPT-6.2 Astragpt-6.2-astra78.1 %config 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-273.3 %config 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
GDP.pdfAA v5.0.1
GPT-6.2 Astragpt-6.2-astra29.8 %config 'GPT-6.2 Astra (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
Claude Fable 5.2claude-fable-5-224.2 %config 'Claude Fable 5.2 (max)' per AAArtificial Analysis Intelligence Index v5.0.1 (independent, same harness across all models)Third-party
FrontierCodev2.0 (Anthropic Nov 2026)
Claude Fable 5.2claude-fable-5-249.1 %xhighAnthropic Fable 5.2 & Mythos 5.2 announcement benchmark table (2026-11-09)Vendor-reported
FrontierSWEv2 (Anthropic Nov 2026)
Claude Fable 5.2claude-fable-5-269.9 %unspecifiedAnthropic Fable 5.2 & Mythos 5.2 announcement benchmark table (2026-11-09)Vendor-reported
OSWorld2.1 (offline set partial, Anthropic Nov 2026)
Claude Fable 5.2claude-fable-5-289.3 %unspecifiedAnthropic Fable 5.2 & Mythos 5.2 announcement benchmark table (2026-11-09)Vendor-reported
Remote Labor Index (RLI)Anthropic Nov 2026
Claude Fable 5.2claude-fable-5-241.3 %highAnthropic Fable 5.2 & Mythos 5.2 announcement benchmark table (2026-11-09)Vendor-reported
Humanity's Last Exam (with tools)Anthropic Nov 2026
Claude Fable 5.2claude-fable-5-274.2 %unspecifiedAnthropic Fable 5.2 & Mythos 5.2 announcement benchmark table (2026-11-09)Vendor-reported

Each row records who published and who ran it (see the Harness column); numbers compare only within a documented run.

Sources: Artificial Analysis (independent) — Independent same-harness comparison: Intelligence Index v5.0.1 (GPT-6.2 Astra, GPT-6.2 Sol, Claude Fable 5.2, Claude Opus 5.5) · Anthropic — Introducing Claude Fable 5.2 and Claude Mythos 5.2 (announcement & benchmark table)

06 // BY WORKLOAD

Which model for which workload

Confidence labels: Measured = same-harness evidence · Price-derived = list-price arithmetic · Inferred = reasoned from docs · Insufficient evidence = no comparable public data.

Recommendations by workload scenario
WorkloadOur pickConfidenceWhy
Quality-firstClaude Fable 5.2MeasuredFable 5.2 (Max) is the #1 overall model in AA v5.0.1 (65 vs 62) with huge leads on AA-Omniscience (67 vs 50), HLE-Diamond (80.2% vs 72.2%), CritPt v2 (76.5% vs 71.5%), and AA-Briefcase (1785 vs 1644).
Agent workflowsGPT-6.2 AstraMeasuredAstra beats Fable Max on both AutomationBench-AA (77.2% vs 76.4%) and Terminal-Bench 4.0 (64.1% vs 59.6%) while completing AA tasks 3.6x faster (11.5 vs 41.1 min) at 1/6.6th the spend ($4.00 vs $26.54).
Complex repository maintenanceDecide by bake-offMeasuredClose split: Fable leads FrontierCode-AA 2.0 (48.8% vs 45.4%, though Astra xhigh reaches 49.0%), while Astra leads shell-heavy Terminal-Bench 4.0 (64.1% vs 59.6%) and scientific repos (78.1% vs 73.3%). Test against your repo budget.
Daily coding tasksGPT-6.2 AstraMeasuredEqual $8/$40 list price turns into $2.07-$4.00/task on Astra vs $9.08-$26.54/task on Fable High/Max because Fable emits far more reasoning output tokens.
Budget-sensitive / high volumeGPT-6.2 AstraMeasuredEvery matched AA effort tier favors Astra by 2.4x to 6.6x on realized per-task spend (e.g., $2.52 vs $14.83 at xhigh for the same Index 62 score).
Long-context tasksClaude Fable 5.2Price-derivedAbove 272K input tokens, Fable stays flat ($8/$40, $0.20 cached reads) while Astra jumps 2x input / 1.5x output ($16/$60, $0.80 cached reads) — Fable is half Astra's input rate on >272K workloads.
Latency-sensitive productionGPT-6.2 AstraMeasuredAlthough Fable streams tokens faster (111 vs 60 tok/s), its heavy reasoning token count pushes mean end-to-end latency on AA Index tasks to 41.1 min (Max) vs Astra's 11.5 min (Max), and Astra adds Ultrafast (3x promo).
Offline batch jobsDecide by bake-offPrice-derivedBoth discount Batch 50% ($4/$20 short); choose Fable Batch when inputs exceed 272K (remains $4/$20 vs Astra Batch $8/$30) or Astra Batch / 8h background (-40%) when output token brevity dominates.

HOLD CRITERIA

When staying on the lower-cost or incumbent model is reasonable

  • Note: both are November 2026 releases; prefer GPT-6.2 Astra when per-task output token explosion would blow your budget ($4.00 vs $26.54 AA Max).
  • Keep GPT-6.2 Astra if your orchestration depends on Ultrafast serving or Responses API -30%/-40% background completion windows.
  • Choose Claude Fable 5.2 only where its +17pt AA-Omniscience edge, #1 reasoning depth (Index 65), or >272K flat pricing outweighs higher output volume.

EVALUATION PROTOCOL

5-step Fable 5.2 vs GPT-6.2 Astra cost-normalized bake-off

Public benchmarks don't run your tasks. Before committing a fleet, measure both models on your own work.

  1. 1Pin output-normalized comparisons: test GPT-6.2 Astra at 'xhigh'/'max' alongside Claude Fable 5.2 at 'medium' ($3.80 AA avg) and 'high' ($9.08), NOT just Max vs Max.
  2. 2Record output tokens AND cached reads on every run — do not estimate monthly bills from the shared $8/$40 list price alone.
  3. 3Segment prompts below and above 272,000 input tokens so Astra's 2x/1.5x cliff enters the model comparison honestly.
  4. 4Check tool_choice constraints before routing production agent loops to Fable 5.2 (avoid 'any'/'tool' when adaptive thinking is on, which is always on in Fable 5.2).
  5. 5Score acceptance cost as (total realized USD across retries) / accepted outputs, not pass rate on a single Max run.

QUESTIONS

Frequently asked questions

How can two models with the exact same $8/$40 price differ by 6.6x per task?

Token pricing only governs rate per token; realized bills equal rate multiplied by generated output tokens. At Max effort on Artificial Analysis v5.0.1, Claude Fable 5.2 outputs far more reasoning tokens per task (41.1 min at 111 tok/s), driving cost per task to $26.54 versus GPT-6.2 Astra's $4.00 (11.5 min at 60 tok/s).

Does Claude Fable 5.2 have any API compatibility gotchas compared to GPT-6.2 Astra?

Yes on both sides: Fable 5.2 always enables adaptive thinking, rejects forced tool_choice ('any' or 'tool') with HTTP 400, and applies 30-day data retention unless authorized. GPT-6.2 Astra requires the Responses API for tool calling, rejects reasoning effort 'none', and changed prefix-preservation behavior.

Which caches more cheaply if 90% of my prompt is a static system/repo prefix?

Below 272K tokens, Fable charges half Astra's cache-read rate ($0.20 vs $0.40/MTok) and identical 5m/30m write fees ($10.00/MTok). Above 272K tokens, Fable is 4x cheaper on cache reads ($0.20 vs $0.80) and half on writes ($10.00 vs $20.00). Watch output verbosity closely via the Cost Calculator.

CONTINUE EXPLORING

VERIFICATION LOG

Sources & freshness

  • Anthropicaccessed 2026-11-18
    Claude Fable 5.2 Overview (model ID, 1M context, limits, retirement floor)https://platform.claude.com/docs/en/models/fable-5-2/overview · published 2026-11-09claude-fable-5-2: 1,000,000 context window, 128,000 max output, knowledge cutoff Jun 2026, minimum retirement 2027-11-09, flat $8/$40 MTok pricing, adaptive thinking low/medium/high/xhigh/max (default 'high').
  • OpenAIaccessed 2026-11-18
    GPT-6.2 Astra Model Documentation (specs, endpoints, effort levels, pricing tiers)https://developers.openai.com/api/docs/models/gpt-6.2-astra · published 2026-11-17gpt-6.2-astra: 1,050,000 context window, 128,000 max output, 272K long-context threshold (2x input/cache, 1.5x output), knowledge cutoff May 18, 2026, reasoning effort low/medium/high/xhigh/max ('none' rejected). Supports Standard, Batch, Fast, Ultrafast (3x promo through 2027-02-28 / 6x starting 2027-03-01), and beta background windows; no Flex tier published.
  • Anthropicaccessed 2026-11-18
    What's new in Claude Fable 5.2 (adaptive thinking, breaking changes, data retention)https://platform.claude.com/docs/en/models/fable-5-2/whats-new-fable-5-2 · published 2026-11-09Documents breaking API changes relative to Opus 5.5 / Fable 5.1: adaptive thinking is always enabled, forced tool_choice ('any' or 'tool') returns 400 error, and 30-day data retention applies by default unless enterprise authorization is granted.
  • OpenAIaccessed 2026-11-18
    OpenAI API Changelog (GPT-6.2 Astra & Sol release, discounted background completion windows)https://developers.openai.com/api/docs/changelog · published 2026-11-17Confirms Nov 17, 2026 launch of gpt-6.2-astra and gpt-6.2-sol alongside 2-hour (-30%) and 8-hour (-40%) discounted background completion windows on the Responses API public beta, plus Responses API prefix-preservation and instructions handling changes in GPT-6.2.
  • OpenAIaccessed 2026-11-18
    OpenAI API pricing (GPT-6 family tier matrix: standard/batch/flex/fast/ultrafast, short/long context, cache)https://developers.openai.com/api/docs/pricingLong context defined as >272K input tokens per request (2x input and cache rates, 1.5x output for the whole request). 'Priority processing' renamed 'Fast mode' on 2026-07-30. Flex and Batch are half of Standard for the GPT-6 family, Fast is 2x, Ultrafast is 6x. Cache writes cost 1.25x the uncached input rate, reads 0.1x (0.05x on GPT-6.1 Sol). Cache-write rates are captured on every published tier row, not Standard only. Re-verified 2026-11-18 for GPT-6.2 Astra ($8/$40 short, $16/$60 >272K) and GPT-6.2 Sol ($2/$10 short, $4/$15 >272K), plus Ultrafast 3x promo through 2027-02-28 (6x list starting 2027-03-01) and absence of Flex rows for GPT-6.2.
  • Anthropicaccessed 2026-11-19
    Claude pricing (API pricing table incl. cache write/read, fast mode, batch)https://claude.com/pricingOpus 5.5: $4/$20 per MTok, cache read $0.20, cache write (5m TTL) $5, batch 'save 50%', fast mode 2x on input/output. Re-verified 2026-11-18: Claude Fable 5.2 ($8/$40, $0.20 cached read = 2.5%, $10.00 5m write).
  • Anthropicaccessed 2026-11-19
    Claude pricing docs (cache TTL prices, fast/batch stacking rules)https://platform.claude.com/docs/en/about-claude/pricingOpus 5.5 1h cache write $8/MTok; caching multipliers apply on top of fast-mode pricing and stack with the Batch discount; fast mode not available with Batch. Re-verified 2026-11-18: Fable 5.2 Batch 50% off ($4/$20), Fast 2x ($16/$80), 2.5% cache-read stacking, 1h cache write $16.00.
  • Artificial Analysis (independent)accessed 2026-11-18
    Independent same-harness comparison: Intelligence Index v5.0.1 (GPT-6.2 Astra, GPT-6.2 Sol, Claude Fable 5.2, Claude Opus 5.5)https://artificialanalysis.ai/models/comparisons/gpt-6-2-astra-vs-claude-fable-5-2 · published 2026-11-18Independent same-harness Artificial Analysis Intelligence Index v5.0.1 suite (also mirrored cleanly across /models/comparisons/gpt-6-2-astra-vs-gpt-6-2-sol, /claude-fable-5-2-vs-claude-opus-5-5, and /gpt-6-2-astra-vs-claude-opus-5-5). Incorporates 9 evaluations: AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset), and Terminal-Bench-Science 0.1.
  • Anthropicaccessed 2026-11-18
    Introducing Claude Fable 5.2 and Claude Mythos 5.2 (announcement & benchmark table)https://www.anthropic.com/claude-fable-and-mythos-5-2 · published 2026-11-09Nov 9, 2026 launch post and official benchmark table comparing Fable 5.2, Opus 5.5, and GPT-6 Astra across FrontierCode 2.0 (49.1% vs 45.5% vs 42.8%), FrontierSWE v2 (69.9% vs 62.3% vs 65.5%), OSWorld 2.1 offline partial (89.3% vs 87.2% vs 79.5%), RLI (41.3% high effort vs 21.3% vs 20.8%), Humanity's Last Exam with tools (74.2% vs 67.7% vs 57.2%), Chartography (73.4% vs 64.4% vs 71.0%), and AA-Briefcase v1.1 (1785 vs 1800 vs 1569 Elo).

Pricing data last verified 2026-11-19. Vendors ship frequently — re-verify before making spend commitments. See the methodology page for update cadence and stale-data flags.