Among the contenders, who will claim the AGI crown?
OpenAI and Anthropic are locked in a fierce battle for dominance in the code field, with the competition never dull.
On the latest Coding ranking of AGI Ranker, Claude Opus 4.8 has just ascended to the throne. GPT - 5.5's score of 77.48 has been outperformed, with a gap of 3.5 points.
Within three months, the programming throne has changed hands several times.
For the first time, Claude Opus 4.7 reached the top, and Anthropic celebrated.
For the second time, GPT - 5.5 overtook, and OpenAI regained the upper hand.
For the third time, Opus 4.8 made a comeback with a score of 81.01, and Anthropic reclaimed the top spot.
Although Anthropic has secretly submitted an IPO application, it suffers from insufficient innate computing power. However, OpenAI and Google DeepMind may still catch up.
Just now, OpenAI has announced a live stream tonight, and Altman may appear.
A battle among the contenders: Will OpenAI draw its sword?
Tomorrow, OpenAI will hold a live event titled "Intelligence at Work" in the morning Eastern Time, and Altman will attend.
Subsequently, Nadella will deliver the opening keynote speech at the Microsoft Build conference in the afternoon Eastern Time.
Microsoft is also expected to release its self - developed new model tomorrow.
Anthropic says it will soon launch a Mythos - level model.
Polymarket predicts a 68% probability that GPT - 5.6 will be released between June 8th and 14th.
This week is crucial for OpenAI.
Will the Codex update go head - to - head with Claude Code?
OpenAI will release a major update to Codex tomorrow.
After months of development, the results are remarkable.
The major update to Codex is not a minor fix.
According to internal sources at OpenAI, the Codex team has been working on this update for months. The core direction is to evolve Codex from a "code completion tool" into a "programming intelligent agent."
Yesterday, OpenAI confirmed that Codex and its cutting - edge models are now fully available on AWS Bedrock, describing it as "the beginning of a broader expansion."
Now, with the imminent release of the major Codex update, this seems to be a coordinated action rather than a coincidence.
Meanwhile, Altman is also promoting the OpenAI Foundation's work on social resilience.
This kind of soft power layout is usually carried out before or after a major release, not during the release process.
There is also an episode: Anthropic secretly submitted an S - 1 filing to the U.S. Securities and Exchange Commission (SEC) yesterday.
Now, both companies are simultaneously racing to advance their IPOs and compete for dominance in the intelligent coding field.
The morning after Anthropic submitted its S - 1 filing, OpenAI announced an important milestone for Codex. This is either an excellent timing or an excellent timing.
The revelation about GPT - 5.6 has caused a stir!
Insiders say the model will be released this week. Its performance is expected to approach the top level of Anthropic Mythos, but at a price 2 - 3 times lower.
Alongside the major upgrade of Codex, there will be a comprehensive evolution in reasoning, front - end generation, personality, and intelligent agent workflows. The efficiency optimization will enable long - process agents to be truly implemented. This could be the most significant AI release in 2026!
- GPT - 5.6 is expected to be released this week, and Codex will also receive a major update.
- It is about 2 - 3 times cheaper but can reach the level of Anthropic Mythos.
- It will bring significant improvements in reasoning, front - end generation, personality, and intelligent agent workflows.
- According to insiders, GPT - 5.6 could easily be called GPT - 6.
- The focus seems to be on efficiency, aiming to make intelligent agents faster, cheaper, and more practical in longer workflows.
The discussion about GPT - 5.6 is not baseless, but there are still several steps before its official release.
It is reported that an entry referencing gpt - 5.6 in the rollout - mapping briefly appeared in OpenAI's internal Codex logs and then disappeared, while most entries pointed to GPT - 5.5.
This is in line with the characteristics of a canary test or a limited internal probe - it is a sign of work in progress, not a release, and certainly not a feature list.
There is a market on Polymarket regarding whether GPT - 5.6 will be released by a certain date.
As of mid - May 2026, traders believe the probability of a public release by June 30, 2026, is approximately 80 - 89%.
This is the public's bet on the time, just a probability, not a commitment from OpenAI. Similar markets have been wrong before.
Looking at the release rhythm: OpenAI quickly released GPT - 5.4, GPT - 5.5, and GPT - 5.5 Instant in succession.
It is a reasonable assumption to continue this rhythm and release a minor version update. "A reasonable assumption based on the rhythm" is the weakest of the three signals - it only shows that GPT - 5.6 is worth looking forward to, but it cannot prove its existence or release time.
However, some netizens revealed that OpenAI is confident and determined to win:
GPT - 5.6 is not a reactive release. OpenAI aims to win, not catch up.
As far as I know, GPT - 5.6 aims to improve by 12% to 15% in some aspects compared to GPT - 5.5. Compared with the average improvement of about 6.5% between Opus 4.7 and 4.8 measured by many, this is a significant step.
GPT - 6 is the real move.
This model not only aims to respond to Anthropic's Mythos series but also to surpass it with a significant advantage.
This is why OpenAI didn't rush to follow up after Opus 4.8.
Anthropic's computing power dilemma: The strongest model, the shallowest pockets
Anthropic's situation is delicate.
It holds top - tier cards.
Opus 4.8 has reached the top of the Coding list, Mythos dominates in the cybersecurity field, and Claude Code's reputation in the developer community continues to rise.
In terms of model capabilities, Anthropic currently has no fear of anyone.
But model capabilities are not the only battlefield.
Anthropic's computing power sources are highly dependent on the cloud infrastructure of Amazon and Google.
It doesn't have an exclusive computing power agreement like Microsoft Azure behind OpenAI, nor does it have the vertical integration advantage of Google DeepMind, where the hardware team is under the same roof.
IPO is a crucial step for Anthropic to make up for this shortcoming. The secret submission of the application has been confirmed, but when the money will arrive, how much will come, and whether it will be enough to keep up with the pace in the computing power arms race are all unknowns.
Anthropic has the strongest model, but not the deepest pockets.
Mythos' pricing strategy also exposes this contradiction.
The Token price, which is 6 times that of Opus, shows that Anthropic is using high pricing to cover high inference costs. This can be sustained in the short term due to its leading capabilities, but once competitors achieve similar capabilities at a lower price, the high - price strategy will become a burden.
Can the IPO provide Anthropic with the computing power ammunition before OpenAI catches up?
This is one of the biggest mysteries in the entire AGI finals.
The time trial of AGI
The throne on the code ability ranking list changes hands on a weekly basis.
This has never happened in AI history.
In past technological competitions, the advantage of the leader was measured in years.
In the search engine field, Google has led for twenty years.
In the mobile operating system, the pattern of iOS and Android has remained essentially unchanged for more than a decade.
But at the forefront of the AGI competition, the gap between the two companies has been