Scan: Will Anthropic be the second-best Code Arena | WebDev AI lab at the end of September 2026?

1. Planner Fable 5.1 · Fermi decomposition + tool plan

Tools chosen: polymarket_direct resolution_source polymarket_related kalshi_related claude_news article_search gdelt_news wikipedia

Sub-questions (Fermi decomposition)

What is the current Lab Rank ordering on the arena.ai Code Arena | WebDev leaderboard (filtered by Labs), and which lab is currently #2?
How large is the Elo/score gap between the top three labs (e.g., Anthropic, Google, OpenAI) on the WebDev leaderboard, and how frequently has the #2 position changed in the past several months?
Which major coding-capable model releases (Claude, Gemini, GPT, Grok) are expected or rumored between now and September 30, 2026, and how have past releases shifted WebDev rankings?
How does arena.ai compute Lab Rank (best model per lab? aggregate?) and has the methodology or leaderboard structure changed recently in ways that could affect resolution?
What prices do the sibling Polymarket markets (Google, OpenAI, xAI as second-best) imply, and do they sum consistently with the Anthropic price?
Historically, has Anthropic tended to hold #1 rather than #2 on WebDev, meaning the 'second-best' outcome depends on another lab overtaking it?

Planner reasoning

This resolves on the Lab Rank of the arena.ai Code Arena WebDev leaderboard on Sept 30, 2026. Key drivers are the current lab ordering (Anthropic vs Google, OpenAI, xAI, etc.), the gap in Elo scores between #1/#2/#3, and expected model releases between now and September that could reshuffle rankings. The Polymarket price and sibling markets (Google/OpenAI/xAI second-best) are the primary anchors.

2. Research Parallel tool calls

tool	status	time	items	summary
polymarket_direct	OK	0.4s	1	## This Market's Polymarket Data Will Anthropic be the second-best Code Arena \| WebDev AI lab at the end of September 2026? - Current price (probability): 71.75% - 7-day price change: -6.25% - 30-day price change: +65.50% - Total volume: $17,856 (USD notional) - Price range: 4.30% - 86.70% - Da
resolution_source	OK	5.8s	1	already_resolved=unclear; matches_criteria=True; last_updated=Sep 11, 2026. The Lab Rank table (filtered for Labs) on the Code Arena \| WebDev Overall leaderboard, dated Sep 11, 2026, shows OpenAI ranked #1 (gpt-6-astra-max, score 1800) and Anthropic ranked #2 (claude-fable-5.1-max, score 1758), with
polymarket_related	OK	1.7s	31	Polymarket search over 4 keyword(s), kept 31 matches. keyword 'Code Arena WebDev': 8 markets \| keyword 'second-best Code Arena': 8 markets \| keyword 'best AI lab end of September': 8 markets \| keyword 'Anthropic Claude': 7 markets
kalshi_related	OK	5.9s	0	0 related markets / summaries. keyword 'LMArena': no matches \| keyword 'Code Arena': no matches \| keyword 'best AI model': no matches
claude_news	OK	31.3s	17	Based on available search results (data is somewhat inconsistent across sources, likely reflecting rapid leaderboard churn in Sept 2026): WebDev/Code Arena standings (model-level, most recent snapshots): - OpenAI's GPT-6 Astra has claimed the top spot on Code Arena's WebDev leaderboard with a
article_search	OK	0.8s	34	Fetched 34 unique articles across 4 queries (lookback=45d). 'LMArena WebDev leaderboard Anthropic Claude ranking': 10 articles \| 'arena.ai Code Arena WebDev top model': 10 articles \| 'Claude vs Gemini vs GPT coding benchmark leaderboard 2026': 10 articles \| 'Anthropic new Claude model release 2026':
gdelt_news	OK	157.7s	10	GDELT: 10 articles across 3 queries (lookback=30d). 'LMArena leaderboard Claude': error GDELT rate-limited after retries (429) \| 'Anthropic Claude coding model release': 10 hits \| 'Gemini coding leaderboard': error GDELT rate-limited after retries (429)
wikipedia	OK	0.3s	3	Fetched 3 Wikipedia entries (0 missing pages).

3. Evidence Brief Sonnet 5 · 7064 chars

# Current state As of the most recent arena.ai snapshot (Sep 11, 2026), the Code Arena | WebDev Lab Rank table shows **OpenAI #1** (gpt-6-astra-max, 1800) and **Anthropic #2** (claude-fable-5.1-max, 1758), a 42-point gap, with Alibaba #3 and Moonshot #4. This matches the "Yes" resolution condition, but the market checks the leaderboard on Sep 30, 2026 12:00 PM ET — three weeks after this snapshot — so the ranking is not yet locked in. # Timeline of key events - 2026-02: Anthropic reportedly held #1, #2, and #4 on the coding/WebDev-style board (reported, KEAR AI via claude_news). - 2026-04: Lab-level summary shows Anthropic leading Text, Code, Document, Search; Google leading Vision/Image/Video (reported, CodeSOTA). - 2026-07-24: Claude Opus 5 released, described as step-change on reasoning/agentic work (confirmed, Anthropic materials via claude_news). - 2026-09-02: Anthropic launches Claude Fable 5.1 and Mythos 5.1, marketed as top coding/knowledge models (confirmed, multiple outlets: CNBC, TechCabal, TOI, Digit, ANI). - 2026-09-03/05: OpenAI launches GPT-6 Astra; tops WebDev leaderboard at 1797, ~35 pts ahead of Claude Fable 5.1 (1762) (confirmed, cryptobriefing.com; corroborated by freepressjournal.in "OpenAI claims lead over Claude Fable 5.1"). - 2026-09-05: arena.ai labs-rank view shows 650,961 votes across 20 labs (reported). - 2026-09-11: Direct resolution-source fetch: OpenAI #1 (1800), Anthropic #2 (1758), Alibaba #3, Moonshot #4, Meta #5 (confirmed, arena.ai direct fetch). # Event Will Anthropic hold the #2 Lab Rank position on the arena.ai Code Arena | WebDev leaderboard (Labs filter) as checked Sep 30, 2026, 12:00 PM ET? # Outcomes to forecast Yes / No # Kalshi market anchor No kalshi_direct price was returned in this research pull (kalshi_related also found 0 matches for "LMArena," "Code Arena," "best AI model"). **Polymarket price is the best available consensus proxy: 71.75% YES**, down 6.25pp over 7 days but up sharply (+65.5pp) over 30 days — reflecting a swing from Anthropic being #1 (bad for this "2nd place" market) toward Anthropic dropping to #2 after OpenAI's GPT-6 Astra launch (good for this market). Price range over the period: 4.3%–86.7%. Volume is thin (~$17.9k total). # Resolution source Fetched https://arena.ai/leaderboard/code/webdev?rankBy=labs — snapshot dated Sep 11, 2026. Currently shows OpenAI #1 (1800), Anthropic #2 (1758), Alibaba #3, Moonshot #4, Meta #5. This matches the "Yes" resolving condition today but is not the final Sep 30 check; leaderboard could shift again before close. # Sub-question answers 1. **Current Lab Rank ordering / #2 lab?** — Per direct Sep 11, 2026 fetch: OpenAI #1, **Anthropic #2**, Alibaba #3, Moonshot #4, Meta #5. 2. **Score gap / volatility of #2 spot?** — Gap between #1 (OpenAI, 1800) and #2 (Anthropic, 1758) is 42 points (~35 pts on Sep 3-5). The #1/#2 order flipped as recently as Sep 3, 2026 when GPT-6 Astra launched and overtook Anthropic, which had been #1 as recently as April 2026 — indicating the top spot is contested and can move quickly with major releases. 3. **Upcoming releases through Sep 30, 2026?** — No confirmed new Gemini beyond 3.1 Pro (felloai.com, swfte.com) or new GPT beyond GPT-6 Astra (Sep 3) found in research. Grok 4.5 entered Agent Arena (Jul 13) but no WebDev-specific listing reported. Anthropic already shipped Fable 5.1/Mythos 5.1 (Sep 2); no further Anthropic release confirmed before close. 4. **Methodology?** — Lab Rank reflects each lab's best-performing model, scored via Bradley-Terry model; WebDev is now a subset of Code Arena/Text Arena per arena.ai's changelog (structural change, but resolution criteria account for this via the "Models" view fallback). 5. **Sibling Polymarket pricing consistency?** — "OpenAI best AI (Sept)" = 73.15%, "Anthropic best AI (Sept)" = 26%, both near-zero for all others — consistent with a two-lab race where OpenAI leads and Anthropic is a strong #2 (71.75% here). Sums aren't required to hit 100% (independent markets) but the directional pattern (OpenAI≈best, Anthropic≈2nd-best) is internally consistent. 6. **Historical #1 vs #2 tendency?** — Anthropic held #1 in Feb and April 2026 lab-level summaries; the #2 position is new as of Sep 2026, contingent on OpenAI's GPT-6 Astra release overtaking it. So "Yes" requires OpenAI to hold on to #1 (or Anthropic reclaim #1, which would flip to No) while Anthropic stays ahead of Alibaba/Moonshot/Qwen challengers. # Key facts (high-confidence, factual) 1. [resolution_source, direct] Sep 11, 2026: OpenAI #1 (1800), Anthropic #2 (1758). 2. [cryptobriefing.com] Sep 3-5, 2026: GPT-6 Astra scored 1797 vs Claude Fable 5.1 at 1762. 3. [CNBC, TechCabal] Sep 2, 2026: Anthropic released Claude Fable 5.1 and Mythos 5.1. 4. [Wikipedia] Anthropic valued at $965B (May 2026), planning IPO; considered rival/peer of OpenAI at frontier. 5. [claude_news/felloai] One conflicting snapshot suggests a Qwen model briefly led model-level WebDev with Anthropic #2 — likely stale or model- vs lab-level confusion; direct resolution-source fetch (dated, primary) is preferred. # Cross-market signals - Kalshi related: none found. - Polymarket (this market): 71.75% YES. - Polymarket siblings: OpenAI "best" Sept = 73.15%, Anthropic "best" Sept = 26%, Google/Meta/Alibaba/Moonshot/ByteDance/Z.ai/SpaceXAI "best" ≈0. October analog: Anthropic 2nd-best = 54.4%, OpenAI 2nd-best = 27% (lower, more uncertain further out). - No sportsbook data applicable. # Analyst opinions and speculation - FreePressJournal (Sep 7, 2026): OpenAI "claims lead" over Claude Fable 5.1 in coding/reasoning — consistent with leaderboard data. - felloai.com snapshot (older, ~2 weeks prior to Sep pull) implies more churn/uncertainty at the very top than the single Sep 11 snapshot suggests. # Directional lean per outcome - **Yes (Anthropic #2):** Supported by current Sep 11 snapshot, Polymarket at 71.75%, Anthropic's strong recent model (Fable 5.1) still #2 by solid margin over #3 (Alibaba). Risk: Anthropic could reclaim #1 (flip to No) or slip below Alibaba/Moonshot/Qwen given China-lab volatility. - **No:** Requires either Anthropic overtaking OpenAI for #1, or a Chinese/other lab (Alibaba, Moonshot, Qwen) leapfrogging Anthropic into #2 before Sep 30 close. Polymarket assigns this only ~28% combined probability. # Gaps / unknowns - No live Kalshi YES price was retrieved for direct comparison; Polymarket used as primary anchor. - Three weeks remain between the Sep 11 snapshot and the Sep 30 check — any new model release (Gemini, GPT, Claude, Chinese labs) could reorder standings. - Conflicting model-level reports (Qwen topping WebDev) create some uncertainty about the exact current live ranking beyond the Sep 11 direct fetch. # Calibration anchors - Polymarket YES price (proxy anchor): 71.75%. - Direct leaderboard snapshot (Sep 11, 2026): matches "Yes" condition today. - Precedent: leaderboard #1 flipped OpenAI↔Anthropic within days of a major model launch (Sep 3, 2026), showing meaningful volatility risk over a 3-week window.

4. Forecasters 5 draws · Fable 5.1 + GPT-5.6 Sol

claude-fable-5-1

58% conf

Yes	73%
No	27%

The direct Sep 11 fetch of the resolution source shows OpenAI #1 (1800) and Anthropic #2 (1758), matching the Yes condition, and Polymarket (the only market proxy available) prices Yes at ~72%. A flip to No requires either Anthropic overtaking OpenAI (a 42-point swing with no further Anthropic release confirmed after Fable 5.1 on Sep 2) or a Chinese lab like Alibaba/Moonshot leapfrogging Anthropic within three weeks, with no confirmed major releases scheduled. The Sep 3 OpenAI↔Anthropic flip shows top-of-board volatility is real, so I keep meaningful tail probability on No, but the status quo over a ~3-week horizon is the most likely outcome and I stay close to the market anchor.

claude-fable-5-1

58% conf

Yes	74%
No	26%

The direct Sep 11 resolution-source fetch shows OpenAI #1 (1800) and Anthropic #2 (1758), matching the Yes condition, and the Polymarket proxy anchor sits at 71.75% YES. Both top labs shipped their flagship models in early September (Fable 5.1 on Sep 2, GPT-6 Astra on Sep 3), making another top-tier release before Sep 30 less likely, while the 42-point gap makes an organic Anthropic reclaim of #1 (which would flip to No) improbable without a new model. The main residual risk is a Chinese lab (Alibaba/Moonshot/Qwen) leapfrogging into #2 or an unexpected Anthropic/Google release over the ~3-week window, plus stale-snapshot uncertainty since the Sep 11 fetch. I lean slightly above the Polymarket price given the status quo already resolves Yes and the short remaining window, but keep meaningful No mass for leaderboard volatility.

claude-fable-5-1

58% conf

Yes	72%
No	28%

The direct Sep 11 resolution-source fetch shows Anthropic #2 (1758) behind OpenAI (1800) with a 42-point gap, matching the Yes condition, and Polymarket (the only market proxy available, no Kalshi price) sits at 71.75% YES. Anthropic already shipped Fable 5.1/Mythos 5.1 on Sep 2 with no further release confirmed before Sep 30, making a reclaim of #1 unlikely absent a new model, while no Chinese-lab or Google release is confirmed that would plausibly leapfrog Anthropic into #2. The main residual risks are unannounced launches (Gemini, Qwen, Moonshot) or Bradley-Terry score drift over three weeks, plus the precedent that the top ordering flipped within days of GPT-6 Astra's launch. I anchor near the market consensus at ~0.72 YES.

gpt-5.6-sol

67% conf

Yes	72%
No	28%

The direct Sep 11 resolution-source snapshot points toward Yes because Anthropic is currently #2 behind OpenAI, while Alibaba and Moonshot trail it. The recent OpenAI-Anthropic flip shows meaningful volatility, but no additional major release from OpenAI, Anthropic, or Google is confirmed before the Sep 30 check. The sibling markets also imply a two-lab hierarchy with OpenAI most likely #1 and Anthropic therefore most likely #2. I remain close to the thin-market Polymarket anchor of 71.75% YES, with residual No risk from Anthropic retaking #1 or another lab leapfrogging it.

gpt-5.6-sol

67% conf

Yes	72%
No	28%

The direct Sep. 11 resolution-source snapshot points toward Yes because Anthropic is currently #2 behind OpenAI. The 42-point OpenAI lead and absence of confirmed major releases before Sep. 30 support the present ordering, while Anthropic's apparent lead over Alibaba and Moonshot reduces the risk of falling below second. However, the recent OpenAI-Anthropic flip demonstrates meaningful release-driven volatility, and Anthropic reclaiming #1 would itself resolve No. I therefore remain close to the thin-market Polymarket anchor of 71.75% YES.

6. Final plain mean of 5 draws

73% (62% confidence)

Yes: 73% No: 27%

Mean of 5 draws (0.73, 0.74, 0.72, 0.72, 0.72). The direct Sep 11 fetch of the resolution source shows OpenAI #1 (1800) and Anthropic #2 (1758), matching the Yes condition, and Polymarket (the only market proxy available) prices Yes at ~72%. A flip to No requires either Anthropic overtaking OpenAI (a 42-point swing with no further Anthropic release confirmed after Fable 5.1 on Sep 2) or a Chinese lab like Alibaba/Moonshot leapfrogging Anthropic within three weeks, with no confirmed major releases scheduled. The Sep 3 OpenAI↔Anthropic flip shows top-of-board volatility is real, so I keep meaningful tail probability on No, but the status quo over a ~3-week horizon is the most likely outcome and I stay close to the market anchor.

Pipeline Timing

Total pipeline time: 246.4s

Per-tool research timings shown in the Research section above.