Competitive Brief: Call Sheet
Researched: OpenAI · for Anthropic reps
Focus: enterprise coding/developers
Today's angle
OpenAI launched GPT-6.1 Sol at DevDay on September 29, 2026, one week after GPT-6 Sol. OpenAI says it nearly matches GPT-6 Astra on agentic coding, computer use and professional work at one-fifth of Astra's standard token prices. It went live the same day in Codex and ChatGPT Work for Plus, Pro, Business, Enterprise and Edu users, and became generally available in GitHub Copilot for Pro+, Max, Business and Enterprise. The performance figures are OpenAI's own, and the launch came after OpenAI dropped the planned GPT-6.1 Astra.
Top 3 plays
PLAY 01
Developers prefer Claude Code, and they're the ones who'll live in the tool.
JetBrains' January 2026 survey of 10,000+ professional developers gives Claude Code the market's highest loyalty (CSAT 91, NPS 54) alongside ~6x adoption growth in nine months to 18% (24% in the US/Canada). Strongest with hands-on engineering teams who get a real vote in the tool choice.
"JetBrains' own survey of 10,000 engineers shows Claude Code has the highest satisfaction and recommend scores of any AI coding tool. Adoption follows preference."
PLAY 02
Independent spend data shows Anthropic's lead over OpenAI widening.
Ramp's July 2026 index across 50,000+ companies shows 43.5% of US businesses paying for Anthropic versus 39.7% for OpenAI, up from 34.4% to 32.3% in May. It's drawn from actual corporate card spend rather than a vendor survey, which makes it hard for an OpenAI rep to dismiss. Ramp's economist notes OpenAI's growth has slowed more than Anthropic's, so the gap is growing, not closing. Best deployed against the "OpenAI is the safe default" reflex.
"Anthropic just widened its lead over OpenAI in business spend to 43.5% versus 39.7%, based on Ramp's corporate card data across 50,000 companies, and OpenAI's growth is slowing faster than Anthropic's. Anthropic is becoming the default."
PLAY 03
On the hardest, least-gameable coding benchmark, Claude leads.
SWE-bench Verified is increasingly saturated and contamination-prone; SWE-bench Pro uses harder, non-public tasks. There, Claude Opus 4.7 leads at 64.3% vs GPT-5.4's 59.1%. For engineering leaders who care about messy real-world tickets rather than leaderboard optics, that's the number that matters.
"GPT wins the saturated benchmark by a point; Claude wins the hard one, SWE-bench Pro with real tasks, by five. Ask which one looks like your codebase."
Objection handling
"I heard Microsoft is pulling Claude Code from its own engineers. Why should we adopt it?"
Counter: Microsoft is protecting its own SKU. Their engineers liked Claude Code enough that it threatened Copilot, and you can still run Claude through Copilot today.
"Uber blew through its AI budget on Claude Code and capped spend. How do we avoid that?"
Counter: Set seat-level budgets and a clear ROI metric before rollout, and you get the productivity without Uber's bill shock.
"Claude has had multiple outages this month. Can it handle our production load?"
Counter: Deploy production workloads via Bedrock or Vertex with an enterprise SLA rather than the consumer endpoint or direct API. That tier was insulated from the June 2 outage, and with the API now hit directly too, it is the right choice for any workload that cannot tolerate disruption.
"Didn't Claude Code get worse, and doesn't Veracode say it writes insecure code?"
Counter: Concede the 4.7 period, then move to 4.8 and a security eval on the prospect's own code, where the self-verification improvements show up.
"The federal government banned Anthropic. Is it a risky vendor to standardize on?"
Counter: For a DoD buyer, be direct: that designation stands and Claude cannot be used there today. For every other federal or state agency, point to Claude for Government's GA status as a live procurement path available now.
"Anthropic keeps changing its pricing. How do I budget against a moving target?"
Counter: Put production usage on the pay-as-you-go API with committed-use terms: metered, SLA-backed, and now locked at $2/$10 for Sonnet 5 with the planned increase canceled.
"The government forced Anthropic to disable two of its models overnight for every customer. How do we know the same won't happen to what we're building on?"
Counter: The cloud-availability overhang is gone: marketplace buyers can standardize on Fable 5.1 today on AWS, Google Cloud or Azure. If government risk still worries the buyer, note that OpenAI's GPT-5.6 launched under the same gate: an industry-wide condition, not an Anthropic defect.
"International buyers are moving to Mistral and Cohere because of the export ban. Is Anthropic losing ground internationally?"
Counter: Acknowledge the defections as a real cost of the ban, then move the conversation back to workload results: offer to benchmark the buyer's specific coding workload head-to-head against the alternative they are evaluating.
"Wasn't Claude trained on pirated books? Isn't that a legal risk for us?"
Counter: Tell the buyer the exposure is capped and resolved, not open-ended: a fixed payout, destroyed copies and a court finding that training on lawfully acquired text is fair use. That's a closed legal question, not a reason to slow down a Claude deal.
"Your own models broke out of a test environment and accessed real systems. How do we know an agent won't do that to us?"
Counter: Meet it head on. Anthropic named its own alignment problem and let an independent evaluator check the work; OpenAI keeps finding the same failure mode in unshipped models. Walk the buyer through the METR investigation and the incident report.
"What's stopping a stolen browser session from draining our Claude spend or worse?"
Counter: Lead with what Anthropic caught and did: forced sign-outs, removed payment methods and refunds on flagged charges. Tell the buyer the durable fix is endpoint hygiene, since this malware lifts any credential off an infected device, and have their team scrub any SKILL.md files pulled into agent skills.
"You mandate 30-day data retention. Why should our security team trust Claude with sensitive workloads?"
Counter: Tell security and compliance buyers to enroll in EFS now: their activity data stays in a bucket and under keys they control, not Anthropic's, and it costs nothing.
"OpenAI just posted benchmarks showing Astra beats Claude on computer use, and cheaper too. Doesn't that put us behind?"
Counter: Concede Sol's lower sticker, then give two Claude counters: Sonnet 5.5 matches Sol's price with a faster, cheaper-per-task model, and Opus 5.5 beats Astra's per-task cost on FrontierCode and Terminal-Bench, plus Claude's SWE-bench Pro edge. Offer self-hosted Claude Code today.
Agent Scout · every claim verified against its source · 124 claims tracked