Claude Fable 5.1 vs GPT-6 Astra Comparison

# Claude Code CLI — an agentic coding session with Fable 5.1
# 1) Install (npm)
npm install -g @anthropic-ai/claude-code

# 2) Pin Fable 5.1 explicitly (aliases can resolve differently depending on environment)
claude --model 'claude-fable-5-1[1m]' --effort high

# 3) Programmatic call via the API (Python)
import anthropic

client = anthropic.Anthropic()
response = client.messages.create(
    model="claude-fable-5-1",
    max_tokens=4096,
    system=[{"type": "text", "text": "You are a senior backend engineer.", "cache_control": {"type": "ephemeral"}}],
    messages=[
        {"role": "user", "content": "Find the N+1 query risk in this Prisma schema."}
    ]
)
print(response.content[0].text)

# 4) Cache economics: a 5-minute cache write costs $12.50/MTok,
#    a cache-hit costs only $0.25/MTok (0.025x multiplier — 0.1x on other models)

# Codex CLI — an agentic coding session with GPT-6 Astra
# 1) Install and update (0.153.4+ makes Astra the bundled default)
npm install -g @openai/codex
codex --version   # must be >= 0.153.4

# 2) Pinning the model explicitly (~/.codex/config.toml — both are top-level keys)
# model = "gpt-6-astra"
# model_reasoning_effort = "high"

codex exec "Find and fix the race condition in this Express route"

# 3) Programmatic call via the API (Python)
from openai import OpenAI

client = OpenAI()
response = client.responses.create(
    model="gpt-6-astra",
    reasoning={"effort": "high"},
    input="Analyze the N+1 risk in this SQL query."
)
print(response.output_text)

# 4) Raising effort without breaking cache (configuration_update)
# add {"type": "configuration_update", "reasoning": {"effort": "xhigh"}} to the input item