{"next_page": 2, "previous_page": null, "max_next_page": 10, "posts": [{"oid": "claude-opus-is-10x-faster-than-openai-gpt-5-at-non-streaming-completions", "title": "Claude Opus is 10x faster than OpenAI GPT 5 at non-streaming completions", "pub_date": "2026-07-24T15:48:18.297Z", "comments": 0, "categories": ["Python", "AI"], "html": "

This picture summarizes it well:

\n

\n

Here on my blog, for this popular blog post I get a lot of comments. 28k blog comments over the years. Some of them are terribly written and hard to understand, so I let AI suggest a rewrite. That code that sends the blog post comment to AI, I actually fire off three times: once with OpenAI gpt-5, once with OpenAI gpt-5-mini, and once with Claude claude-opus-4.8. I use my human eyes and judgement to evaluate the results, and I can tell you they do equally well. Only the slightest differences.

\n

The surprising thing is how amazingly slow OpenAI's gpt-5 is! It's nearly 10x slower than claude-opus-4.8. What's up with that!?

\n

It's also clear that the latency difference between gpt-5-mini and gpt-5 is significant. At the time of writing, the input token price difference between gpt-5.4 and gpt-5.4-mini is $2.50 compared to $0.75! That's a 3x difference.

\n

Conclusion

\n\n

Bonus

\n

Before I added Claude, I used to use litellm to wrap OpenAI's models. The code looks like this:

\n
\nresponse = litellm.completion(\n    model="openai-gpt-5",\n    api_key=settings.OPENAI_API_KEY,\n    messages=my_prompt_messages,\n)\n
\n\n

Unlike, if you use the native OpenAI Python SDK the invocation looks like this:

\n
\nclient = openai.OpenAI(api_key=settings.OPENAI_API_KEY)\nresponse = client.responses.create(\n    model="gpt-5,\n    input=my_prompt_messages,\n)\n
\n\n

I measured the difference, in speed, where I compare using the OpenAI SDK versus the litellm wrapper and the difference looks like this:

\n

\n

Granted, in June I \"only\" did a bit over 30 of these calls, but strangely there's a difference!
\nI don't have the intricate knowledge to understand why the litellm makes the total time different from using the SDK. (not sure I care either!)

\n

Either way, I'm moving away from litellm and only use the SDKs provided by Claude and OpenAI. Feels safer given the CVEs we've seen this year on litellm.

", "url": null, "disallow_comments": false, "split": null}, {"oid": "best-django-redis-configuration", "title": "Best Django Redis configuration", "pub_date": "2026-07-20T19:11:05.415Z", "comments": 0, "categories": [], "html": "

Latest blog post about different compressors for Django Redis.

", "url": "https://www.peterbe.com/plog/best-django-redis-configuration-for-speed-and-size", "disallow_comments": false, "split": null}, {"oid": "best-django-redis-configuration-for-speed-and-size", "title": "Best Django Redis configuration for speed and size", "pub_date": "2026-07-19T20:01:27.016Z", "comments": 0, "categories": ["Django", "Python"], "html": "

In 2017 I wrote \"Fastest Redis configuration for Django\". Now, I've made an update for 2026 that focuses on comparing default Redis with three compressors:

\n\n

The results on macOS are:

\n
                        TIMES        AVERAGE   MEDIAN (P50)   MEDIAN (P90)         STDDEV\ndefault                    198        0.181ms        0.176ms        0.216ms        0.038ms\nzlib                       213        0.190ms        0.182ms        0.237ms        0.039ms\nlzma                       134        0.273ms        0.272ms        0.332ms        0.042ms\nzstd                       202        0.192ms        0.180ms        0.243ms        0.046ms\n\nBest Means (shorter better)\n###############################################################################\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                        0.181  default\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                      0.190  zlib\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588  0.273  lzma\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                     0.192  zstd\n\nBest Medians (shorter better)\n###############################################################################\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                         0.176  default\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                       0.182  zlib\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588  0.272  lzma\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                        0.180  zstd\n\nSize of Data Saved (shorter better)\n###############################################################################\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588  5151  default\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                                        2151  zlib\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                                                 1420  lzma\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                                          2010  zstd\n\nSize of Data without Default (shorter better)\n###############################################################################\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588  2151  zlib\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588                         1420  lzma\n\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588       2010  zstd\n
\n

The \"conclusion\", on macOS, is that lzma is slowest but compresses the best. On Ubuntu (in GitHub Actions), the fastest of the compressors is zstd and the best compression is lzma.

\n

But it's a very small difference.

\n

The type of data you store in Redis will matter for your use case. In this experiment, it's storing strings of numbers. Each number is about 10 characters long.

", "url": null, "disallow_comments": false, "split": null}, {"oid": "how-to-use-a-listtuplearray-in-django-with-a-raw-sql-cursor", "title": "How to use a list/tuple/array in Django with a raw SQL cursor", "pub_date": "2026-07-14T13:14:40.383Z", "comments": 0, "categories": ["Django", "Python"], "html": "

This does not work:

\n
\nfrom django.db import connection\n\nlist_of_values = [1, 2, 3]\nwith connection.cursor() as cursor:\n    cursor.execute("""\n        SELECT *\n        FROM my_model_table\n        WHERE some_value IN %s\n    """, [\n        tuple(list_of_values),\n    ])\n    results = cursor.fetchall()\n
\n\n

It will give you:

\n
django.db.utils.ProgrammingError: syntax error at or near "'(1,2,3)'"\nLINE 4:         WHERE id IN '(1,2,3)'
\n\n

It used to work with psycopg v2. Now, in psycopg v3, you have to use the ANY operator. See \"You cannot use IN %s with a tuple\"

\n

This will work:

\n
\nfrom django.db import connection\n\nlist_of_values = [1, 2, 3]\nwith connection.cursor() as cursor:\n    cursor.execute(\n        """\n        SELECT *\n        FROM my_model_table\n        WHERE some_value = ANY(%s)\n    """,\n        [\n            list_of_values,\n        ],\n    )\n    results = cursor.fetchall()\n
\n\n

Note the ANY(%s), and instead of a list that has a tuple, it's a list that has a list.

\n

What About a List of Strings

\n

Consider...

\n
\nfrom django.db import connection\n\n-list_of_values = [1, 2, 3]\n+list_of_values = ['foo', 'bar', 'fiz']\nwith connection.cursor() as cursor:\n    cursor.execute(\n        """\n        SELECT *\n        FROM my_model_table\n        WHERE some_value = ANY(%s)\n    """,\n        [\n            list_of_values,\n        ],\n    )\n    results = cursor.fetchall()\n
\n\n

That will result in:

\n
django.db.utils.DataError: invalid input syntax for type integer: "foo"\nLINE 4:         WHERE some_value = ANY('{foo,bar,fiz}')
\n\n

My solution was to rewrite the SQL string itself and treat each value as a parameter each. In other words, the SQL string, before being sent to cursor.execute(...) will contain something like this:

\n
\nAND (\n  some_value = % OR\n  some_value = % OR\n  some_value = % OR\n  some_value = % OR\n  -- ...etc...\n  some_value = %\n)\n
\n\n

This will work and is safe:

\n
\nfrom django.db import connection\n\nlist_of_values = ["foo", "bar", "fiz"]\nwith connection.cursor() as cursor:\n    cursor.execute(\n        f"""\n        SELECT *\n        FROM my_model_table\n        WHERE ({" OR ".join(["some_value = %s" for _ in list_of_values])})\n    """,\n        list_of_values,\n    )\n    results = cursor.fetchall()\n
", "url": null, "disallow_comments": false, "split": null}, {"oid": "driving-range-and-thunder-clouds", "title": "Driving range and thunder clouds", "pub_date": "2026-07-12T10:26:11.412Z", "comments": 0, "categories": [], "html": "", "url": null, "disallow_comments": false, "split": null}, {"oid": "4th-of-july-tablecloth-that-we-all-contributed-to", "title": "4th of July table cloth that we all contributed to", "pub_date": "2026-07-04T22:19:09.620Z", "comments": 0, "categories": [], "html": "", "url": null, "disallow_comments": false, "split": null}, {"oid": "potpourri-of-funly-cut-hot-dogs", "title": "Potpourri of fun\u2019ly cut hot dogs", "pub_date": "2026-07-04T20:28:04.320Z", "comments": 0, "categories": [], "html": "", "url": null, "disallow_comments": false, "split": null}, {"oid": "putted-in-my-first-ever-eagle", "title": "Putted in my first ever Eagle!", "pub_date": "2026-06-28T17:18:19.306Z", "comments": 0, "categories": [], "html": "", "url": null, "disallow_comments": false, "split": null}, {"oid": "celebratory-1776-titleist", "title": "Celebratory 1776 Titleist", "pub_date": "2026-06-28T17:15:25.102Z", "comments": 0, "categories": [], "html": "

Unfortunately, I hooked this one into the water on the seventh hole.

", "url": null, "disallow_comments": false, "split": null}, {"oid": "a-little-praying-mantis-climbing-on-the-screen-on-my-back-porch", "title": "A little praying mantis climbing on the screen on my back porch", "pub_date": "2026-06-25T00:07:45.134Z", "comments": 0, "categories": [], "html": "", "url": null, "disallow_comments": false, "split": null}]}