Historical source material recovered on 2026-09-09. Capture dates and source limitations apply; recovery is not fresh verification.

Agentic Model Economics

Cost-performance trade-offs for selecting models in tool-using and long-running agent workflows.

Key points

• Added to AI Watch from ^[src-191].

Related

• Claude Sonnet 5
• AI Coding Agents

2026-07-07 buyer-side cost pressure update

• The Register's AI bubble discussion connects model economics to infrastructure capital risk: if enterprise adoption and AI revenue fail to justify hyperscaler build-out, the correction affects more than model vendors ^[src-209].
• Electrek's Tesla report adds an operator-level cost signal: employee token consumption can become material enough to trigger weekly caps and model-routing policy ^[src-210].

Source references

• ^[src-191] Anthropic – "Introducing Claude Sonnet 5" (2026-06-30)
• ^[src-209] Brandon Vigliarolo / The Register – "Even banks and hyperscalers are now sounding the alarm about the AI bubble" (2026-07-06)
• ^[src-210] Fred Lambert / Electrek – "Tesla caps employee AI spending at $200/week except for Grok" (2026-07-02)

2026-07-10 AI Watch batch update

• New source evidence added: ^[src-211], ^[src-214], ^[src-223], ^[src-224], ^[src-225], ^[src-227], ^[src-231]. See the paired source summaries for verified claims and evidence caveats.

Historical recovered snapshot — 2026-08-03

Preserved from pr-123. Documentary recovery on 2026-09-09; claims below retain their historical source scope and were not freshly verified. Current synthesis above remains authoritative.

• OpenAI cut Luna and Terra sticker prices while charging a premium for Sol Fast throughput, making price, intelligence and latency independently selectable dimensions ^[src-270].
• For agents, lower per-token rates can be outweighed by longer outputs, extra reasoning steps, tool calls or retries; benchmark task completion and total run cost together ^[src-270].

Evidence checked on 14 September 2026

The Infographics Show presents a critical financial narrative about compute commitments, operating losses, funding dependence, GPU useful-life assumptions and an advertising-led revenue strategy. ^[src-445] This is secondary financial commentary, not an independently verified audit. Do not carry the narrated numbers or allegations as audited facts. The narration mixes historical figures, forecasts, cash burn, net loss and noncash charges and contains internal inconsistencies; no numerical financial claim was validated in this video review.

• The Leaked Audit That Exposes OpenAI's Real Financial Crisis