Skip to main content
AI CANADA PULSE
ExploreDated · source-linked

Cohere & the frontier

A Canadian voice. A global frontier.

Follow model providers and the very different kinds of capital behind them.

Sources checked 25 September 2026

AI-generated concept illustration · not a photograph

The capital context

Amounts need context.

Closed financing, commitments and corporate spending
are different things. They are not added together.

MethodSources checked 25 September 2026 · conversions use June 5, 2026 reference rates. Each row is a different kind of capital.

Structured / upcoming commitment

Cohere / Aleph Alpha / Schwarz Group

EUR500M structured financing commitment for Cohere's upcoming Series E.

What this includes

Shown for context, but not counted as a closed private round or added to Cohere's benchmark.

No valuation disclosed in the announcement.

C$804MCanadian-dollar reference
Committed capital

OpenAI

US$122B committed capital announced Mar. 31, 2026.

What this includes

Committed capital scale; not comparable to public grants, procurement, or cloud-credit support.

US$852B post-money valuation.

C$170BCanadian-dollar reference
Closed private financing

xAI

US$20B Series E announced Jan. 6, 2026.

What this includes

Series E financing is the scale comparison. No recent valuation is shown unless a cited source discloses it.

No source-supported recent valuation is wired for this row.

C$27.8BCanadian-dollar reference
Public-company capex

Meta

US$125B-US$145B 2026 capital expenditure guidance, including finance leases.

What this includes

Meta is not a private financing round. It is included as a frontier-model peer with a public-company infrastructure spending signal.

Approx. US$1.52T public market capitalization snapshot as of June 7, 2026.

C$174B-C$202BCanadian-dollar reference

The model landscape

Go straight to the providers.

Provider source links, without
unapproved benchmark scores.

Independent frontier benchmark

How long can models work alone?

METR measures the length of tasks a model can complete with 50% success on its task suite. It is a capability signal, not a forecast of jobs, general intelligence, or Canadian economic impact.

Capability doubling time129 daysEstimated doubling time for the 50% task-time horizon since 2023METR time horizons (opens in a new tab)
Latest model trackedClaude Mythos Preview Early50% time horizon ~ 17 h (released 2026-04-07)METR time horizons (opens in a new tab)
Models benchmarked26METR time-horizon suite, GPT-2 era to todayMETR time horizons (opens in a new tab)
GPT-53.4 h
Released 2025-08
Gemini 3 Pro3.7 h
Released 2025-11
Claude Opus 4.54.9 h
Released 2025-11
GPT-5.25.9 h
Released 2025-12
Claude Opus 4.612 h
Released 2026-02
Claude Mythos Preview Early17 h
Released 2026-04

Live METR TH1.1METR warns that measurements above 16 hours are unreliable with the current task suite. The benchmark is concentrated in software engineering, machine learning, and cybersecurity tasks; it does not represent all intellectual work or jobs.