GPT-Rosalind vs Qwen 4 Max: 3000 agentes, una sola factura y un modelo que aún no se ha lanzado

Thirty-two minutes, and more than 700 references. On 21 October 2026 O​penAI published a collaboration with PerturbAI and Scripps Research in which thousands of agents running on GPT-Rosalind worked through a whole-brain CRISPR atlas of millions of mouse cells, each agent taking one gene disruption, searching the literature around it, and writing a report — a job the authors estimate at months to a year done one perturbation at a time. The model itself is not new: O​penAI introduced GPT-Rosalind on 16 April 2026, widened it beyond a named partner list on 11 September, and began billing at $5.00 per million input tokens and $25.00 per million output on 5 October. What is new is the first public look at the model running as a fleet rather than as a call, and that changes what a comparison with Q​wen 4 Max should be about. Q​wen 4 Max was announced at A​libaba's Apsara conference on 22 September 2026 and still has no price, no context window and no model card. So the honest version of this matchup has three sides: a specialist whose real unit cost is per agent, an opponent that has not shipped, and the Q​wen model you can actually rent today — Qwen3.8-Max, at $2.00/$6.00 per million tokens. Announcements that never become a callable endpoint are recorded with their dates and sources in the AI Incident Archive, alongside the rest of the model and vendor incidents we follow.

Lo que realmente mostró la ejecución del 21 de octubre

El conjunto de datos es un atlas de Perturb-seq: genes alterados de a una guía de ARN por vez en todo el cerebro del ratón, con el transcriptoma de cada célula leído después. El resultado es una matriz de miles de perturbaciones frente a millones de células, muchas más de las que cualquier laboratorio puede examinar en detalle, que es el cuello de botella que aborda el artículo.

The method, which the authors call distributed biological reasoning, is not one large model call over a big table. Each agent receives a single gene disruption, examines the cellular response, retrieves the surrounding literature, and produces a structured report proposing a mechanism. Those reports are then assembled into what the paper calls a reasoning atlas: a graph linking genes, cell-type effects, papers, and candidate mechanisms. The runtime O​penAI reports for the whole sweep is 32 minutes.

El hallazgo más importante para quien elija un modelo no es la biología. Es el control que el equipo aplicó a su propio método, porque ese sí tiene una cifra asociada. Mantuvieron fijos el modelo y la capacidad de cálculo y dieron a un único agente el conjunto de datos completo en lugar de repartirlo entre miles. A medida que crecía la carga de trabajo, la cobertura bibliográfica del agente único se desplomó: con la mayor carga de trabajo citó 11 referencias frente a más de 700 de los agentes distribuidos. Los expertos humanos enmascarados prefirieron los informes distribuidos el 75% de las veces.

Read those together and the result is about decomposition, not about model IQ. A single context window degrades when you pour an atlas into it; a few thousand narrow agents with a retrieval tool each do not. The authors' own framing supports this — they present the DCTN1 cholesterol-transport hypothesis as a question for follow-up experiments, not a discovery — and the paper is a preprint posted on 16 October 2026, co-authored by O​penAI and PerturbAI employees, with no peer review yet. Treat the 32 minutes as a real runtime and the human comparison as an order of magnitude.

Para una página de comparación, la consecuencia es concreta. Si la victoria provino de la arquitectura, entonces la propiedad decisiva del modelo subyacente no es su puntuación en benchmarks — es cuánto cuesta invocarlo tres mil veces en media hora, y si tienes permiso para hacerlo.

El número que decide este enfrentamiento es el costo por agente.

Published comparisons of these two models tend to line up headline rates and stop. Headline rates are the wrong unit for the workload the 21 October paper just made concrete, because a distributed run multiplies the model call by the number of perturbations. Work it through with stated assumptions: an agent that reads a shared atlas preamble, pulls a handful of papers, and writes a report might consume 60,000 input tokens and 4,000 output tokens. At O​penAI's published rates that agent costs $0.40 — $0.30 of input and $0.10 of output. The same agent on Qwen3.8-Max at $2.00/$6.00 costs $0.14. Three thousand agents: about $1,200 against about $432.

Cache the shared prefix and the gap widens rather than closes, because the cached-input rates diverge more than the headline ones. Put 50,000 of those 60,000 input tokens in a cached prefix and the Rosalind agent drops to roughly $0.18 while the Q​wen agent drops to roughly $0.06 — a little over three times cheaper, because cached input is $0.50 per million on one side and $0.25 on the other. That is our arithmetic, not a vendor figure, and it is sensitive to how much of your prompt is genuinely shared; the direction is not.

Ponlos frente a frente en las dimensiones que determinan una ejecución de flota:

• Input price — GPT-Rosalind $5.00 per 1M vs Qwen3.8-Max $2.00. Q​wen 4 Max: not published.

• Output price — GPT-Rosalind $25.00 per 1M vs Qwen3.8-Max $6.00. Q​wen 4 Max: not published.

• Cached input — GPT-Rosalind $0.50 per 1M, and O​penAI's pricing page states cache-write pricing does not apply to the model; Qwen3.8-Max $0.25. Q​wen 4 Max: not published.

• Cost per 3,000-agent sweep — about $1,200 on GPT-Rosalind vs about $432 on Qwen3.8-Max at the assumptions above. Q​wen 4 Max: not computable.

• Context — 1M tokens on both sides. Q​wen 4 Max: not published, and specifically not inherited from the Qwen3.8 window by virtue of the tier name continuing.

• Salida máxima — GPT-Rosalind no publica un límite por respuesta; la instantánea Qwen3.8-Max 0902 documenta 131.072 tokens. • Latencia — Qwen3.8-Max mide un tiempo hasta el primer token p50 de 6,32 segundos en OrcaRouter y 40,3 tokens de salida por segundo en el índice Artificial Analysis. GPT-Rosalind no publica ninguna cifra de latencia, que para una ejecución de fan-out es el dato que más querrías.

• Puntuación independiente — GPT-Rosalind no aparece en absoluto en el Índice de Inteligencia de Artificial Analysis, ya que es un modelo de dominio con acceso restringido; Qwen3.8-Max registra 45, ocupa el puesto 33 de 214 modelos, a 5,41 $ por tarea de índice completada y 40,3 tokens de salida por segundo.

• Access — GPT-Rosalind requires clearance through O​penAI's trusted-access programme; Qwen3.8-Max is a normal commercial API with an open sign-up.

Two of those lines do more work than the rest. The independent score is the only number in the table nobody involved in either model supplied, and it says the Q​wen flagship is a competent mid-frontier generalist rather than a frontier leader — 45 against 58 for the best model on that board. And the access line is not a footnote: it decides whether the fleet can exist at all.

Where Q​wen 4 Max actually stands

A​libaba used its Apsara conference on 22 September 2026 to name four tiers — Q​wen 4 Max as the flagship, Q​wen 4 Flash for throughput, Q​wen 4 Plus in the middle, and an open-weight Q​wen 4 27B — described the family as training on a new-generation architecture, and said it would arrive soon. It attached no date to any of them.

A month later there are three independent places a shipped Q​wen flagship would be visible, and it is absent from all three. A​libaba Cloud's own Model Studio documentation lists qwen3.8-max as the highest Max tier, with no Q​wen 4 entry. Artificial Analysis, which scores most of the commercial Chinese flagships within days of release, has no Q​wen 4 Max row. The OrcaRouter catalogue, which carries more than 200 models including every Q​wen 3.5 through 3.8 tier, has no qwen4 identifier. That is not proof of a delay. It is proof that as of 22 October there is nothing to compare.

The parameter figure that travelled with the announcement deserves one correction, because it keeps propagating into comparison tables. The 5-trillion-to-10-trillion number in circulation is a forward-looking statement about the generations after Q​wen 4 — the Q​wen 4.5 and Q​wen 5 timeframe. A​libaba has attached no parameter count to Q​wen 4 Max. A Q​wen 5-class size printed next to a Q​wen 4 model name is a borrowing error with a long half-life.

What the announcement does tell you is that A​libaba intends to keep contesting the flagship tier, which is a reason to stay inside the family rather than leave it. It tells you nothing you can plan a migration around, because "soon" is not a date and a conference slide is not a rate card.

Qué compra la prima de especialista y qué no

GPT-Rosalind's case rests on domain evaluations O​penAI published with the model: leading BixBench performance, six of eleven LABBench2 wins against GPT-5.4, and per-token gains of 53.7% on GeneBench, 18.0% on MedChemBench and 19.6% on LabWorkBench. Every one of those is vendor-reported. None has been reproduced independently, and the model has no entry on the general intelligence boards — which is normal for a gated domain model and also means the numbers cannot be checked from outside.

Lo que es comprobable es el conjunto de herramientas. El Life Sciences Research Plugin conecta el modelo con más de 50 herramientas científicas y fuentes de datos como habilidades componibles dentro de Codex, y el artículo del 21 de octubre es una demostración de ese stack más que de razonamiento puro: miles de agentes, cada uno de los cuales recupera literatura y escribe un informe estructurado, es un resultado de integración de herramientas. Qwen3.8-Max no tiene un stack de biología equivalente. Tiene una sólida capacidad general de llamada de herramientas y salidas estructuradas, y cada conexión con una base de datos científica tendrías que construirla tú.

Esa es la verdadera naturaleza del sobreprecio. No pagas 2,5× por un modelo más inteligente; lo pagas por un razonamiento ajustado al dominio sobre evidencia molecular y por una capa de herramientas preintegrada, y la ejecución del 21 de octubre es la primera evidencia pública de que la combinación produce algo que un biólogo leerá. El contrapunto incómodo está en el mismo artículo: como el experimento de control mantuvo fijo el modelo y solo cambió la descomposición, el resultado en gran medida se mantendría con un modelo más barato. El estudio no aísla cuánto de la salida corresponde a GPT-Rosalind y cuánto a la arquitectura, y los autores no afirman que lo haga. Si estás presupuestando un barrido distribuido, esa ambigüedad vale unos $770 por cada tres mil agentes, que es la diferencia entre las dos facturas anteriores.

La puerta es parte de la especificación.

GPT-Rosalind is not generally purchasable. O​penAI's pricing page states that access is limited to approved internal research through the trusted-access programme, and since 11 September that programme has been open globally to eligible organisations rather than to a fixed partner list. Eligibility is assessed on the proposed use, governance and controlled access — which is a reasonable policy for a model that reasons over pathogen and protein data, and a hard constraint on the fleet pattern the company just demonstrated.

Una ejecución de 32 minutos con miles de agentes supone que puedes escalar bajo demanda a miles de solicitudes concurrentes. Con una clave sujeta a aprobación, la limitación no son los tokens por minuto, sino la revisión, la estructura de la cuenta y el límite de concurrencia que conceda el programa. Cualquier equipo que planifique un barrido distribuido debería tenerlo en cuenta antes de comparar tarifas, porque un modelo más barato al que puedes llamar libremente supera a uno mejor al que solo puedes llamar según el calendario de otra persona.

This is where a routing layer earns its place, and where it does not. GPT-Rosalind is not on OrcaRouter — access runs through O​penAI's trusted-access application, and no router changes that. What a router does change is everything around the specialist call. A run like the one published on 21 October is mostly not biology: it is orchestration, report parsing, deduplication across thousands of outputs, and the code that turns an atlas into something queryable. Those steps do not need a $25-per-million-output domain model, and putting them there is how a 32-minute experiment becomes an expensive one. Qwen3.8-Max is on OrcaRouter at Alibaba's $2.00/$6.00 list price with 0% markup, behind the same OpenAI-compatible endpoint as more than 200 other models, so the bulk leg and the specialist leg sit on one key with automatic failover between providers and the routing policy written as configuration rather than as a model name baked into the application.

When Q​wen 4 Max ships, that is also the cheapest way to find out whether it matters: a routing rule that sends a slice of traffic to the new tier, with failover back to the model you already trust, is a trial that cannot take production down. Until then the catalogue simply has no entry for it, which is what an announced-and-unreleased model should look like.

¿Qué haría que esta comparación se invirtiera?

Four published numbers would turn Q​wen 4 Max from a name into an opponent, and they are worth naming so the real thing is recognisable when it appears.

• The rate card. A Q​wen flagship that undercuts $2.00/$6.00 changes the arithmetic in this article by a factor rather than a percentage. One that matches it changes nothing, because the buyable Q​wen model already sits there.

• Comportamiento con contexto largo, expresado como política. Si la ventana de 1M conlleva una banda de recargo y a partir de qué umbral. Una tarifa escalonada por encima de cierto tamaño de entrada importaría más en una ejecución de atlas que la cifra titular, porque cada agente de ese patrón cruza la banda.

• An agentic and tool-use result on a board nobody at A​libaba runs. The gap this article describes is not general reasoning — Qwen3.8-Max is competent there — it is scientific tool use. A published, independently administered result on that axis would be the first real test of the specialist premium.

• Whether the Max tier ships weights. Qwen3.8-Max did not, but the 27B tier of the same generation did, and a Q​wen 4 Max with downloadable weights would be the one thing GPT-Rosalind structurally cannot offer a regulated lab that needs to audit what happens to proprietary sequence data.

La decisión que esta página realmente respalda no espera a ninguno de ellos. Coloque al especialista donde equivocarse sale caro —priorización de objetivos, propuestas de mecanismos, el informe que leerá un biólogo— y ponga a la flota en el modelo que cuesta $0.14 por agente y responde sin formulario de solicitud. El resultado del 21 de octubre es un buen argumento para ejecutar biología distribuida a escala atlas. Es un argumento más débil para ejecutarlo todo en el modelo más caro de la sala, y el propio experimento de control del artículo es la razón para decirlo.